Sei sulla pagina 1di 17

Ships and Offshore Structures

ISSN: 1744-5302 (Print) 1754-212X (Online) Journal homepage: http://www.tandfonline.com/loi/tsos20

Logistics-based fleet design for complex


transportation scenarios

Tomaso Gaggero, Giuliano Vernengo, Marco Parodi & Enrico Rizzuto

To cite this article: Tomaso Gaggero, Giuliano Vernengo, Marco Parodi & Enrico Rizzuto (2018):
Logistics-based fleet design for complex transportation scenarios, Ships and Offshore Structures,
DOI: 10.1080/17445302.2018.1454095

To link to this article: https://doi.org/10.1080/17445302.2018.1454095

Published online: 27 Mar 2018.

Submit your article to this journal

Article views: 17

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tsos20
SHIPS AND OFFSHORE STRUCTURES, 
https://doi.org/./..

Logistics-based fleet design for complex transportation scenarios


a a a b
Tomaso Gaggero , Giuliano Vernengo , Marco Parodi and Enrico Rizzuto
a
Department of Electric, Electronic, Telecommunication Engineering and Naval Architecture, University of Genova, Genova, Italy; b Department of
Industrial Engineering (DII), University of Naples Federico II, Napoli, Italy

ABSTRACT ARTICLE HISTORY


A new integrated method for the preliminary design of a fleet of ships for a new trade is presented. Data from Received  August 
two apparently separate types of analysis, based on logistics and naval architecture, are merged within an Accepted  March 
integrated logistics-fleet design framework. The procedure carries out a preliminary screening of the design
KEYWORDS
space based on logistics criteria. Then, possible fleet configurations and ship designs are investigated, pro- Simulation-based design
viding information to shipyards and ship owners for an effective selection of optimal solutions. Interactions (SBD); logistics; fleet
between logistics evaluations and classic ship design assessments are highlighted. Results from a realistic synthesis model (FSM);
test case concerning compressed natural gas transportation are presented and discussed. compressed natural gas
(CNG); design of experiments
(DoE); decision-making
support

1. Introduction ship, as e.g. in Papanikolaou (2010), relations among the units


The relation between the performances of a ship and those of the composing a fleet are investigated. This latter subject can have
fleet including the ship is often ignored at the design stage. The a different impact depending on the type of trade. Consider
analyses of the two aspects are mainly carried out in series (logis- e.g. a trade that aims at transporting a certain amount of goods
tics coming later) and by experts from different fields, namely per year from a loading to an unloading terminal. If there are
naval architects and logisticians. This means that the feed-back no technical limitations on the number of ships allowed at both
from logistics is rarely converted into design actions and, when it terminals and if there is an infinite availability of goods at the
occurs, the study of possible modifications is likely to introduce loading terminal, the interaction among the elements of the fleet
great difficulties and delays in the global design process. is minimal. Introducing a few constraints into this ideal sce-
In addition, both ship and fleet design are characterised by nario, the mutual interactions among the ships of the fleet rise,
interactions of so many quantities covering, e.g. technical, eco- creating more complex situations. Suppose e.g. that the load-
nomic, logistic, safety and many more aspects, that an increased ing/unloading pier at one or both the terminals allows only a
level of analysis is anyway required to account for all interac- ship per time. Then if two or more ships arrive simultaneously
tions and reduce uncertainties as the design process evolves. In at the same terminal, one will be operating, while the other ones
this very complex framework, a simulation-based design (SBD) will have to wait. In this case, considering a perfectly stationary
approach offers the chance of exhaustively exploring wide and system, departures and arrivals of the various ships at a given
multiple design spaces at different levels of fidelity, providing terminal may be distributed in time so that no overlap occurs.
information about the feasibility and quality of the designed In a real system, however, perturbations will occur to the
solutions. To this aim, a holistic approach able to account for timetable of the single ships, such as delays due to adverse
all the possible connections and to discover the sensitivity of weather conditions, faults or accidents (see e.g. Spanos et al.
the input parameters with respect to the design choices repre- 2008 or Papanikolaou et al. 2014) and in port operations or cargo
sents the right path to create effective modelling tools. A new handling (see, for instance, Wang and Meng 2012). Such per-
SBD framework is proposed to accelerate the design process of turbations would lead again to the possibility of having more
the whole trade system (including both ship and fleet design), than one ship ready to dock at the same pier. Such a situation
merging a fleet synthesis model (FSM) with a preliminary logis- strongly affects the choice of the optimal fleet, having both tech-
tics model. nical and economic consequences. Indeed, a ship waiting idle
The technical design of the ships of a fleet, performed by the for some time does not contribute to the transportation process
FSM, is developed on the basis of the information provided by but her operative costs continue to run in that time frame. The
the logistics model, which is responsible for the initial screen- discipline taking into account this type of relationships among
ing of the trade scenario. This integrated design process is then the units of a fleet is exactly logistics. In this field, lots of anal-
able to explore automatically a multidimensional design space yses have been carried out, especially regarding liner container
defined by parameters from both fields. ships (see, for instance, Hsu and Hsieh 2007; Imai et al. 2009;
In addition to the exploration of the interactions of various Lin and Tsai 2014; Karsten et al. 2015). In those works, however,
technical and economic aspects within the design of a single ships have often been considered as black-boxes, without mod-

CONTACT Giuliano Vernengo giuliano.vernengo@unige.it


©  Informa UK Limited, trading as Taylor & Francis Group
2 T. GAGGERO ET AL.

Figure . Map of the Carribean Sea (downloaded from the website https://freevectormaps.com). (This figure is available in colour online.)

elling the existing peculiar and complex interactions between 16

the systems on board. By means of the new approach, an inte-


14
gration of the transportation logistics with the technical design
of the ships of a fleet at a very early design phase is made possi- 12 PIPELINE
Gas delivery [Bcm/year]

ble. Works dealing with a similar logistic-based approach have


10 LNG
been proposed by Brett et al. (2006) and by Boulougouris et al.
(2012) applying the approach to a conventional trade realised by 8
LNG - GTL
a fleet of ro-ro vessels. The authors, however, maintain a system-
level perspective, avoiding the discussion of extensive result. To 6

demonstrate the benefits of the proposed design framework, an 4


application case that is particularly suited to exploit the interac- CNG
Stranded
tions between the logistics model and the FSM is examined, i.e. 2

the design of a fleet of compressed natural gas (CNG) carriers in


0
a realistic scenario. 0 1000 2000 3000 4000 5000
Distance to Market [km]

Figure . Ranges of application of the major technologies for natural gas transport
2. Case study as function of the distance-to-market and the gas delivery. CNG bounds are high-
The application example is represented by the selection of the lighted by bold lines.
main characteristics of fleets of ships supplying CNG to sev-
eral islands in the Carribean Sea (see the map in Figure 1). In pipelines, at least in a range of specific combinations of distance
a marine context, the transportation of NG by ships in the form and gas quantities, as shown in Figure 2 (see Economides and
of compressed gas (CNG) is a relatively new solution, represent- Wood 2009). According to the figure, the most promising sit-
ing a competitive alternative to the shipping of liquefied nat- uations correspond to short distances and relatively small gas
ural gas (LNG) or gas to liquid (GTL) or to the adoption of quantities. The concept is still at a research stage, as, at present,
SHIPS AND OFFSHORE STRUCTURES 3

there are very few examples of operative CNG ships and most Table . Technical constraints to the CNG trade.
of them operate over very short routes or even in inland waters. Type Value
Investigations and feasibility studies carried out in the last years
have, however, prompted studies on possible solutions to the Volume – (MMscf)
Loading/offloading rates – (MMscf/day)
main technological issues related to this type of ships (see, for Distance – (nM)
instance, Wagner and van Wagensveld 2002; Valsgård et al. 2004; Ship speed < (knots)
Lothe 2005; Stenning et al. 2012). The solution has also been
covered by normative requirements. The selected scenario for
the analysis corresponds to the supply of NG to a group of 17 Table . Input/output data to/from the logistics model.
islands in the Caribbean sea. The loading terminal is located in Data Symbol Units
the twin island country of Trinidad and Tobago.
The choice of this situation is motivated both by the real- INPUT SCENARIO
Adjacency Matrix AM Nm
ism of the case (actual possibility of establishing such a type of Consumption rates qC MMscf/h
marine transport) and by the complexity of the scenario, repre- Terminal connection time tC h
sented by the number of possible routes connecting the loading INPUT FLEET
Ship speed (variable range) VS kn
terminal to the receiving ones. The NG supply is meant to be Number of ships (variable range) nS –
the only (or at least one of the main) form of energy supply to Offloading rate (fixed) qOL MMscf/h
the islands, so that the service needs to be continuous, i.e. the gas OUTPUT
Ship speed VS kn
demand needs to be satisfied without interruptions. In principle, Number of ships nS –
this can be achieved placing a storage in each island and organ- Ship capacity 
Q MMscf
S
ising a re-fuelling by ships. However, in view of a containment Distribution patterns DPs –
of costs, a hypothesis can be to avoid the construction of fixed
structures and use the ships moored at ports as storage them-
selves, providing a continuous supply of gas. In other words, in
per fleet nS and the nominal ship speed VS (i.e. the design speed).
this case, a ship must be moored at each terminal at any time.
Values for the main parameters are taken in the range proposed
This peculiar situation makes this case of marine transport par-
by Dunlop and White (2003) (see Table 1) in order to ensure the
ticularly suitable to test the effects of logistics criteria in design-
technical feasibility of the possible designed fleets.
ing a new fleet. The costs arising from other possible installa-
tions on land that support the CNG distribution have not been
considered. 3. Preliminary logistics model for feasibility analysis
The scenario is defined by: According to the formulation proposed by Nikolaou (2010), a
simple and effective logistics model has been implemented in
r the Adjacency Matrix AM (see Appendix 1), containing all the SBD framework. The logistics model is used to obtain a set
the relative distances between the destinations in nautical of fleets (each made by identical ships) able to perform the trade,
miles i.e. able to guarantee a constant offloading ratio for each desti-
r the consumption rates qC in (MMscf/h) required by each nation for a given complex scenario (a gas producer and a set
terminal (see Appendix 2) of gas consumers). In particular, the model requires a few input
r the terminal connection time tC in hours, i.e. the time spent parameters reported in Table 2 together with the provided out-
by each ship to manoeuvre to reach the quay and start dis- puts. The logistics model performs a systematic variation of the
charging the gas (considered equal to 3 h) free variables (VS and nS ) and evaluates which combination can
reach the goal of ensuring the delivery of gas at each destina-
The logistics model, which represents the first step of the tion at each time. Each identified combination is characterised
global design process, aims at the definition of the constraints by a number of ships nS , a nominal capacity of each ship QS and
in terms of ships characteristics able to guarantee the above- a nominal speed V S . An interesting aspect of this investigation
mentioned condition of continuity in the service. As regards the regards the distribution pattern (DP). Two main DPs are taken in
ship design in a narrow sense, an FSM for design and optimi- consideration, namely the Hub-and-Spoke (HAS) and the Milk-
sation of fleet of CNG ships has been developed and tested in Run (MKR), whose schemes are shown in Figure 3.
different idealised scenarios, as in Vernengo and Rizzuto (2014), The HAS scheme is a two-point pattern from a source to a ter-
Vernengo et al. (2016) and Gaiotti et al. (2016). Each candidate minal back and forth and is associated to the above-mentioned
fleet, featuring different dimensions, speed and number of ships, criterion of not using any storage. One ship is always moored at
is designed on the basis of a required gas capacity and evaluated the unloading terminal while the other ones are sailing (accord-
both in terms of technical and economic performances. This ingly, the minimum number of vessels for the HAS is 2). Such
process generates a multi-criteria decision-making (MCDM) kind of DP is particularly suited for destinations either very far
environment that allows for the choice of the optimal fleet based from the hub or with a very large gas request.
on a given target, expressed as a combination of technical and On the contrary, in the MKR, two or more offloading termi-
economic performances defined by the designers. nals in sequence starting from the same source are covered. This
In the next sections, the integrated approach is described in DP is particularly suited for those destinations either close to
detail and results from the application case are shown in the last the hub or with a limited gas request. The other distinctive fea-
section. The fleets are characterised by the offloading rate qOL ture of the two DPs is that the MKR requires a storage at each
(MMscf/h) and two free variables such as the number of ships offloading terminal in order to ensure a suitable gas reservoir,
4 T. GAGGERO ET AL.

Figure . Scheme for Hub-and-Spoke (left side) and Milk-Run (right side) distribution patterns.

while in the HAS the ship itself can be seen as a temporary stor- connection times at each NP , according to Equation (6).
age while discharging the gas directly into the supply natural gas
grid. The final output of the model is therefore a fleet configura- QS QS
TMKR = + + TN , (5)
tion together with a DP for a given scenario. qOL qLO
Ltrip
TN = 2Tc (NP + 1) + . (6)
VS
3.1. Hub-and-Spoke formulation
From the perspective of the various ports, the consumption time
Considering a HAS scenario, the time TRT required by a ship
Tcons can be defined based on the overall consumption rate qc =
to perform the whole round trip, including the loading phase
qc1 + qc2 + · · · + qcNP according to Equation (7). The minimum
and the connections time Tc , and the time TOL needed by all the
gas capacity of the fleet involved in the MKR is defined in Equa-
other ships of the fleet to offload the gas are provided by Equa-
tion (8), that is found imposing TMKR ࣘ Tcons .
tions (1) and (2), respectively. The capacity QS of a ship of the
fleet is found imposing TOL ࣙ TRT , resulting in Equation (3). The QS
total capacity of the fleet QFleet − HAS is simply computed multi- Tcons = , (7)
qc
plying by the number of ships as in Equation (4).
nS TN qc
QFleet−MKR ≥ . (8)
nS − qqOLc − qc
qLO
Ltrip QS
TRT = 4Tc + 2 + , (1) 3.3. Logistics results example analysis
VS qLO
In the following, the results of a systematic logistic analysis
QS
TOL = (nS − 1) , (2) regarding the Jamaica island (within the global Caribbean Sea
qOL Scenario) are shown in Figure 4. The Jamaica island is served
T · qOL following a HAS distribution pattern due to the combination of
QS ≥ , (3)
nS − 1 − qqOLL the required gas capacity and the distance from the source ter-
minal. This is an example set of intermediate results that will be
QFleet−HAS = nS · QS . (4)
further processed in the proposed integrated approach by the
FSM.
The influence of four main design parameters on the ship
3.2. Milk-Run formulation nominal capacity QS is represented in Figure 4. The design
When an MKR pattern is chosen, the rationale is different. In parameters analysed are the ship speed VS , the number of ships
this case, storage facilities are installed at each receiving port NP . per fleet nS , the offloading gas capacity qOL and the percentage of
Each nS vessels involved in the MKR unload the gas at each NP the required amount of energy from fossil sources at the destina-
receiving ports, visited in succession. A fraction of that unloaded tion port QR , defining the required gas that needs to be delivered
gas is stored so that the storage is used before the next vessel to the island. For the present analysis, the actual energy demand
arrives. The arrival must occur before the whole amount stored in terms of fossil sources of each destination is evaluated accord-
is used. From these considerations, the total time TMKR of a cycli- ing to CIA (2016) (see Appendix 2). Four percentages of QR are
cal MKR is the sum of the time spent for loading and unload- considered, 25%, 50%, 75% and 100%: the latter case means that
ing operations plus that needed to complete the round trip, as the whole energy amount from fossil sources needed at the des-
defined in Equation (5). TN is the total travelling time plus the tination island is supplied by natural gas.
SHIPS AND OFFSHORE STRUCTURES 5

30 30
nS=2 - QOL=50% nS=3 - QOL=50%
nS=2 - QOL=100% nS=4 - QOL=50%
28 28 nS=2 - QOL=100%
nS=3 - QOL=100%
26 26 nS=4 - QOL=100%

24 24

22 22
QS [MMscm]

QS [MMscm]
20 20

18 18

16 16

14 14

12 12

10 10

8 8
10 12 14 16 18 20 10 12 14 16 18 20
VS VS

(a) QR = 25%. (b) QR = 50%.

30 30
nS=3 - QOL=50% nS=4 - QOL=50%
nS=4 - QOL=50% nS=5 - QOL=50%
28 nS=5 - QOL=50% 28 nS=6 - QOL=50%
nS=7 - QOL=50%
nS=6 - QOL=50% nS=8 - QOL=50%
26 nS=3 - QOL=100% 26 nS=3 - QOL=100%
nS=4 - QOL=100% nS=4 - QOL=100%
nS=5 - QOL=100% nS=5 - QOL=100%
24 24 nS=6 - QOL=100%
nS=7 - QOL=100%
22 22
QS [MMscm]

QS [MMscm]

20 20

18 18

16 16

14 14

12 12

10 10

8 8
10 12 14 16 18 20 10 12 14 16 18 20
VS VS

(c) QR = 75%. (d) QR = 100%.


Figure . Example of combinations of feasible fleets for a given Hub-and-Spoke scenario. Two offloading capacities qOL are shown on each plot. The required capacity QS
for each ship of a fleet is shown versus the ship speed VS . Required capacity at the offloading terminal QR at %, %, %, % of the maximum attainable is shown from
the top left plot clockwise. (a) QR = %; (b) QR = %; (c) QR = %; (d) QR = %.

Looking at Figure 4, it can be seen that as QR increases, i.e. for smaller as nS increases, meaning that the rate of the improve-
larger amount of gas delivered, solutions made of a higher num- ment due to this design parameter features an asymptotic trend.
ber of ships per fleet appear in the feasibility domain. Coher- This indirectly highlights that there must be a balance between
ently, at a fixed QR , the range of the ship speed VS decreases as the reduction in total CapEx obtained with larger ships and the
nS increases. Setting qOL as high as possible, i.e. less time devoted increment in the total capacity of the fleet needed because a ship
to discharge, (continuous dotted curves) allows to have solu- is always moored at port and not participating in the transporta-
tions with a lower number of ships and in general with a lower tion process. From this viewpoint, it is better to have a small ship
QS . Considering that the Capital Expenditure (CapEx ), i.e. the idling than larger ones. Finding the balance point is a non-trivial
investment costs, represents the main cost of this kind of ships, task because most of the steps of both the ship design process
reducing QS directly reflects in more economic designs. The and the fleet cost estimation introduce strong nonlinear rela-
other favourable effect in this respect is obtained when working tions, making impossible to find a solution of this problem in
in the range of higher speeds. Higher VS again means reduced a sort of closed form.
QS (in the context of the same scenario) at the cost of higher Considering that in this proposed scenario the loading and
Operational Expenditure (OpEx ). Since these costs, remaining offloading rates are fixed, the number of ships has a direct influ-
within reasonable limits, are significantly lower with respect to ence on the loading and offloading time of the fleets. If a fleet is
the CapEx , the overall costs of the fleets are lowered. made of a single ship, during the operations at the terminals, the
In Figure 4(d), there is also a significant gap in terms of QS whole fleet is not sailing, reducing the efficiency of the system.
between the curves of the fleets with nS = 3 and nS = 4 at qOL = Increasing the number of ships results in higher efficiency of the
100%. Such distance between curves at subsequent nS becomes transport, since, in principle, there is a higher percentage of the
6 T. GAGGERO ET AL.

Table . Input/output data to/from the FSM. nominal value. Such a reduction occurs when the ships of the
Input Output fleets sail in rough sea conditions. Within the FSM, the speed
reduction is statistically estimated considering the response of
, V , n )
(Q Pressure vessels
S S S the ship to a specific scatter diagram. The information about the
DP Weights and stability
CM, SM Powering and consumption
reduction of speed with respect to the design speed (VS ) is par-
L, B, D Costs (CapEx, OpEx) ticularly important for the scope of the present work. Such speed
reduction will have a negative influence on the gas delivery of the
whole fleet as the synchronisation among the ships (which is a
hypothesis at the basis of Nikolaou 2010 method) is no more
fleet sailing while some ships are stopped, loading and offload-
respected. The amount of such effect cannot be directly quan-
ing the cargo. Such a dependency is stronger for fleets made of
tified by the FSM without performing simulations in the time
few ships, reflecting in higher slopes of the curves with higher
domain of the entire trade.
nS in Figure 4.
Each designed fleet is then ranked both on the basis of the
The above-analysed results, in particular those presented in
average reduced speed VR , and on the basis of its economic
one of the charts of Figure 4 at fixed QR , represent the typical
characteristics including the breakdown of fixed and variable
output of the logistics model. As it can be noted, many different
costs, the CapEx and the OpEx , respectively. In particular, VR
feasible solutions are available. It is clear that further analyses are
is the weighted average speed maintained by a ship under nor-
necessary in order to make a proper selection among the feasible
mal load and considering all the possible combinations of mean
ones. To this aim, the feasible fleets are designed by means of the
period and significant wave height for a given sea zone with
FSM. The FSM gives preliminary design of each feasible fleet
the corresponding probability of occurrence. A complete defi-
making an estimation of costs and efficiency of each solution in
nition of the various parts of the core of the FSM approach can
terms of effectively delivered gas. Details on the FSM are given
be found in Vernengo and Rizzuto (2014). The computation of
in the following chapter.
VR is described in detail in Vernengo et al. (2016).

4. Fleet synthesis model


5. Integrated logistics-fleet design framework
As mentioned, the FSM allows for the generation and for the
Figure 5 shows the flowchart of the integrated method for auto-
technical and economical evaluation of many fleet variations.
matic fleet design. As described in Section 3, given a set of
Table 3 reports for the FSM the sets of input and output
fixed parameters (orange block in Figure 5) defined by the anal-
parameters, the latter ones described as general categories. Each
ysed scenario and a first sub-set of free design variables com-
designed fleet is characterised by a specific number of ships, ship
posed by combinations of ship design speed VS and number of
design speed, dimensions, Capacity Margin (CM) and Sea Mar-
ships nS (first green block), the logistics model (blue block) per-
gin (SM). Regarding the last two design parameters, the former
forms a preliminary screening providing all the feasible fleets
one represents the percent increase in capacity with respect to
and the related DPs (red block ‘Feasible Fleets’). These pieces
the nominal one which is computed with nominal velocity. The
of information are passed to the Fleet Synthesis Model (blue
latter is the increase in installed power with respect to the engine
block) that designs the fleets, using as free design variables: ships
power needed to reach the nominal speed in still water condi-
dimensions, CM (0%–10% ) and SM (10%–30%). One of the
tion. They have been introduced in order to let the FSM span-
main hypotheses on which this logistic approach is developed
ning over a wider range of possible solutions. In fact, accord-
is that the ship speed is kept constant all over the path. This is
ing to the typical design spiral process (see, for instance, Taggart
equivalent to consider the ships sailing in calm water, which is
1980), the nominal design capacity does not need in principle a
clearly far from the real usual operational conditions. This draw-
margin coefficient, since it is fixed by several boundary condi-
back of the preliminary logistics model is overcome by the FSM
tions of the trade (e.g. the relative distances, ship speed, number
in which, according to the procedure outlined above, the ship
of ships of the fleet and others). In the common practice, the SM
speed reduction due to rough sea conditions is accounted for.
is usually considered by increasing the calm water resistance of
the hull by 15%–20%. The proposed numerical design frame-
work allows for automatic generation and comparison of thou- 6. Test case results
sands of solutions in a very short time. Due to the relatively high The proposed integrated logistics-FSM framework is tested on
level of uncertainty in the context of a preliminary assessment a real scenario, i.e. the design of a new CNG ship-based supply
of a new fleet (the process is indeed meant to provide feasible system covering the islands of the Carribean Sea. The design of
solutions representing an input to develop preliminary designs such a complex system has been performed following the steps
of the ships), using capacity and SM variations allows to inves- shown in Figure 5 and described in Section 5.
tigate beyond the conventional used boundaries. In fact, it has Results from the logistics model are reported in Table 4. A
been demonstrated (see, for instance, Vernengo et al. 2016) that specific DP is assigned to each island based on the combination
using a combination of CM and SM provides the possibility to of the distance from the source terminal (taken from the Adja-
find and hence to evaluate and possibly choose feasible solutions cency Matrix) and the consumption rate qC of the island itself.
that minimise the costs, while obtaining the requested amount Since both the design ship speed VS and the number of ships per
of cargo delivered per year. fleet nS are variables in most of the cases, the same service can
The FSM also provides the fundamental information about be performed by different fleets (i.e. a different combination of
the average reduction of the ship speed compared to the nS and VS ). Due to the combination of qC and distance, some of
SHIPS AND OFFSHORE STRUCTURES 7

Figure . Flowchart of the integrated method for automatic fleet design. (This figure is available in colour online.)

Table . Results of the preliminary logistics analysis: feasible distri- Table . Variation ranges for the free variables in the fleet
bution patterns (DPs) and combinations of nS and VS for all the desti- synthesis model (FSM).
nations of the scenario.
Design variable Symbol Range
Island DP nS VS (kn)
Sea Margin SM %–%
Aruba HAS   Capacity Margin CM %–%
Curacao HAS  –, V =  Longitudinal number of PVs nX –
Jamaica HAS –, nS =   Lateral number of PVs nY –
Jamaica HAS –, nS =  
Jamaica HAS –, nS =  
Jamaica HAS –, nS =  
Jamaica HAS –, nS =  
Jamaica HAS –, nS =   Each combination listed in Table 4 is then evaluated by the
Jamaica HAS –, nS =  
All the other islands MKR –, nS =   FSM that scans all the possible feasible fleets according to vari-
All the other islands MKR –, nS =   ations of the parameters reported in Table 5.
(including Aruba) Considering the range of variations of the design variables,
All the other islands MKR –, nS =  
All the other islands MKR –, nS =   thousand fleets are generated per DP. On average, a fleet takes
All the other islands MKR –, nS =   about 10 seconds to be created and evaluated; the whole fleet
All the other islands MKR –, nS =   design process runs for about 5 days on a single cpu.
All the other islands MKR –, nS =  
Due to the huge amount of data created by the design process,
results are discussed in the following in terms of the main effects
of the design variables instead of looking at each single destina-
the initial destinations (in particular, Haiti, Dominican Repub- tion. Most of the results are displayed as 2D scatter plots in the
lic and Bahamas) lay out of the boundaries traced in Figure 2 plane defined by the reduced speed VR , in the abscissa, and by
and hence have been excluded in the proposed analysis. Another the fleet costs, in the ordinate. The latter quantity might be both
relevant aspect is that Aruba is always served as a HAS destina- the total fleet costs and one of the costs breakdown items. The
tion except when considering the MKR2 in which this port is fleets shown in the results have already been filtered according
included. The other MKR distribution pattern (named MKR1), to the general criterion on the CM expressed in Equation (9) for
which does not include Aruba, is realised at VS = 12 kn for nS = the HAS and the MKR patterns. Depending on VR , the crite-
[5; 11]. The possible DPs resulting from the preliminary logistics ria provide the lower limit for the CM below which a fleet, even
model are also shown in Figure 6 where HAS patterns are repre- if technically possible, is not coherent with the results obtained
sented by blue arrows, whereas the ports involved in MKR1 and from the logistics analysis. Complete derivation of the coeffi-
MKR2 are connected by black and green arrows, respectively; cients α = α(qOL , qL , nS , Tc , QS ), β = β(qOL , qL , nS , Ltrip , QS ), γ =
the two Milk-Runs are also travelled in opposite directions. γ (n, qc , qOL , qLO , NP , QS ) and φ = φ(Tc , qc , NP , Ltrip , QS ) is given
8 T. GAGGERO ET AL.

Figure . Possible distribution patterns configuration for the complete trade system. The loading terminal, i.e. Trinidad and Tobago island, is highlighted in red. HAS possible
patterns are shown by blue arrows. Ports belonging to MKR and MKR are connected by black and green arrows respectively. Background map is taken from the website
https://freevectormaps.com. (This figure is available in colour online.)

in Appendix 3. not including Aruba). The fleet and storage costs are shown in
 Figure 8 against the reduced speed VR for different nS ; the effect
CM ≥ α + VβR , for HAS found by the logistics model in terms of ship nominal capacity
(9) (see, for instance, Figure 4) is confirmed by the FSM, i.e. there
CM ≥ γ + VφR , for MKR
is an inverse dependency between the overall fleet costs and nS ;
The influence of the SM and the CM can be shown analysing consistently, this trend reaches a limit for which increasing the
Figure 7 for the Aruba HAS in the case corresponding to fixed number of ships does not result in lower costs. In particular, the
VS = 12 kn and nS = 2 case. As expected, Figure 7(a) shows improvement (in terms of costs reduction) between two sets of
that the lower speed reduction is experienced by those fleets solutions is higher for steps of nS in the lower range of values
having the higher SM (magenta dots). Nevertheless, fleets with (e.g. between nS = 5 and nS = 6); such a rate of improvement
lower SM (e.g. black and green dots) present slightly lower decreases while nS increases; solutions for nS = 10 and nS = 11
reduced speeds VR but also at lower fleet costs. The same fleets of are practically equivalent in terms of fleet and storage costs. The
Figure 7(a) are also shown in Figure 7(b) but the different CMs CapEx breakdowns for three fleets of the final MKR1 solution
are highlighted with colours. The CM causes a vertical shift of (VS = 12 Kn) at nS = 5, nS = 8 and nS = 11, respectively, are
the points, meaning that it does not significantly affect the speed shown in Figure 9. Table 6 reports the main characteristic of the
reduction while having a strong impact on the costs, particularly three selected fleets. As expected, the dimensions of the ships
in terms of CapEx. Consistently with the limits posed by Equa- as well as the number of pressure vessels (i.e. vessels designed
tion (9), there are no fleets with CM = 1 since it always happens to contain NG compressed at high pressures) on board decrease
that VR < VS . as the number of ships per fleet rises due to the increased fre-
The impact of the number of ships per fleet nS can be under- quency of arrivals/departures. As a direct consequence of the
stood from the MKR1 distribution pattern at V = 12 Kn (the one smaller dimensions, the cost of the steel for the first ship, i.e. the
SHIPS AND OFFSHORE STRUCTURES 9

×10 6
1.4
SM=10
SM=15
SM=20
1.35 SM=25
SM=30
Fleet Cost [k$]

1.3

1.25

1.2

1.15
12 11.9 11.8 11.7 11.6 11.5 11.4 11.3 11.2
Reduced speed [Kn]
(a) Sea Margin effect.

×10 6
1.4
CM=1.025
CM=1.05
CM=1.075
1.35 CM=1.1
Fleet Cost [k$]

1.3

1.25

1.2

1.15
12 11.9 11.8 11.7 11.6 11.5 11.4 11.3 11.2
Reduced speed [Kn]
(b) Capacity Margin effect.

Figure . Fleet costs vs. actual reduced speed VR for the Aruba HAS trade. Effects of Sea Margin SM and Capacity Margin CM are highlighted in the upper and lower sub-
figures, respectively. (a) Sea Margin effect; (b) Capacity Margin effect. (This figure is available in colour online.)

prototype representing the fleet, decreases too. Costs for the rate compared to the rate of increasing of the number of ships.
entire fleets feature a non-monotonic trend with respect to the This is more evident in Figure 9 comparing the fleet made of 11
number of ships. This is due to the nonlinear relations estab- ships with that made of 8 ships. The former, in fact, has higher
lished among the ship dimensions, the costs of the single ship global costs for the steel with respect to the latter one, despite
and the number of ships, as already highlighted in Section 4. the fact that smaller ships are considered. Such behaviour is not
Such a non-monotonic behaviour is due to the fact that the costs present in the costs for PVs and storage, which monotonically
reduction achieved by using smaller ships decreases at a lower decrease as the number of ships increases. This effect occurs

Table . Characteristics of the selected fleets shown in Figure .

nS VS L B D SteelCship SteelCf nPVs PVCship PVCf


(–) (kn) (m) (m) (m) (k$) (k$) (–) (k$) (k$)

  . . . . .  . .


  . . . . .  . .
  . . . . .  . .
10 T. GAGGERO ET AL.

×10 7
1.3
nMR=5
nMR=6
Fleet & Storage Cost [k$] 1.2 nMR=7
nMR=8
nMR=9
1.1
nMR=10
nMR=11
1

0.9

0.8

0.7
12 11.9 11.8 11.7 11.6 11.5 11.4 11.3 11.2
Average reduced speed [kn]

Figure . Fleet costs vs. actual speed for the MKR distribution at the design speed of  kn. Effects of the number of ships per fleet. (This figure is available in colour online.)

because the efficiency in the transport rises (less capacity is increases. Machinery costs are more dependent on the num-
idling when the ship is at port), while the savings in terms of ber of ships rather than on the request of the single ship, hence
hull weight are not so evident when designing a huge number of increasing with the number of ships. The costs for the pres-
small ships for which the ratio between hull weight and payload sure vessels and the storage (the latter being made of pressure

×10 5 Steel ×10 4 Machinery


3
6
Machinery Costs [k$]
Steel Costs [k$]

2
4

1 2

0 0
5 8 11 5 8 11
Number of ships Number of ships

×10 6 Pressure Vessels ×10 6 Storage


6
2.5
Pressure Vessels Costs [k$]

Storage Costs [k$]

2
4
1.5

1
2
0.5

0 0
5 8 11 5 8 11
Number of ships Number of ships

Figure . CapEx breakdown for three fleets belonging to the final MK distribution pattern at the same speed VS =  kn with nS = , nS =  and nS = .
SHIPS AND OFFSHORE STRUCTURES 11

×10 6
7
nHAS =4
nHAS =5
6.5
nHAS =6
nHAS =7
6 nHAS =8
Fleet Cost [k$]

5.5

4.5

4
18 17 16 15 14 13 12 11
Reduced Speed [kn]

Figure . Fleet costs against actual reduced speed VR for the Jamaica Island HAS. (This figure is available in colour online.)

vessels too) are greater than the steel costs of an order of magni- HAS). Another relevant characteristic of these scenarios is that
tude and of two orders of magnitude compared to the machin- the spread among the solutions in terms of fleet costs narrows
ery costs. Consequently, the total CapEx and storage costs for for increasing ship speed. This is due to the combined effect of
the fleet assumes the same monotonic decreasing behaviour of CapEx and OpEx which are detailed in Figure 12 for the MKR2.
the pressure vessels and storage costs. The CapEx decrease for increasing nS and VS since both con-
The last two analysed scenarios correspond to the Jamaica tribute to design fleets with smaller ships, but the OpEx (right
HAS and the MKR2, shown in Figures 10 and 11, respectively. figure) present a clear inversion of the trends, as nS = 11 is
Both cases are much more complicated compared to the pre- reached. Considering growing VS , for 6 ࣘ nS ࣘ 11, the OpEx of
viously proposed analysis, since all the design variables assume the fleet drop, while for nS ࣙ 12, they start rising. This effect is
a range of values, creating a complex multi-dimensional design due to the contrasting trends that are show in Figure 13 between
space where it is hard to take a decision without a proper sup- the costs of the consumables and those of the other running
port tool. In both cases, not all the combinations of number of costs, both contributing to the overall fleet OpEx (refer to Ver-
ships and ship speed produce possible fleets: in fact, at lower nengo and Rizzuto (2014) for a detailed definition of each cost
VS , there are more solutions since higher nS are allowed (e.g. contribution); consistently, the consumption gets larger with
fleets with nS = 7 are found for VS ࣘ 14 kn in the Jamaica higher VS since the request propulsion power is higher while the

×10 7
1.6

1.5
nMR=6
Fleet & Storage Cost [k$]

1.4
nMR=7
nMR=8
1.3
nMR=9
1.2 nMR=10
nMR=11
1.1 nMR=12
nMR=13
1 nMR=14

0.9

0.8
18 17 16 15 14 13 12
Average reduced speed [kn]

Figure . Fleet and storage costs for the MK. (This figure is available in colour online.)
12 T. GAGGERO ET AL.

×10 6
12

11
nMR=6
10 nMR=7
nMR=8
CapEx [k$]

9 nMR=9
nMR=10
8 nMR=11
nMR=12

7 nMR=13
nMR=14

5
18 17 16 15 14 13 12
Reduced speed [kn]
(a) CapEx .

×10 6
4.2

4
nMR=6
nMR=7
3.8 nMR=8
20·OpEx [k$]

nMR=9
3.6 nMR=10
nMR=11
nMR=12
3.4
nMR=13
nMR=14
3.2

3
18 17 16 15 14 13 12
Reduced speed [kn]
(b) OpEx .

Figure . Fixed costs vs. reduced speed for the MKR. CapEx (top) and OpEx (bottom). (a) CapEx ; (b) OpEx . (This figure is available in colour online.)

running costs decrease because they are directly proportional to The final choice is not unique and can be made based on
the size of the ships (the latter is in fact reducing); since the con- the specific needs of each trade. In order to provide an exam-
sumable costs are of a lower order of magnitude with respect to ple of complete results, for each DP, the fleet able to get closer to
the running costs (because of the low cost used for the fuel), the the nominal speed at the lowest possible cost has been selected.
trend of the latter is the predominant one in the final OpEx ; this Details of these fleets are reported in Table 7. In this case, the
is the main reason why the OpEx show the same behaviour of choice of MKR2 and the HAS patterns for Jamaica and Curacao
the CapEx . results in the lowest overall fleet costs of the trade, about 11.81 ×
As described above, the integrated FSM provides information 106 k$ against 13.76 × 106 k$ of the other option, corresponding
about all the possible fleets for each combination (VS and nS ) of to MKR1 and the HAS patterns for Jamaica, Aruba and Curacao.
a specific DP, resulting in a wide and complex multidimensional The ships of the MKR2 and Jamaica HAS fleets have comparable
scenario. Considering that the cpu time needed to design a fleet dimensions, while the Curacao HAS is realised by slightly larger
ranges in the order of few seconds (from less than one up to three ships. The nominal speed VS is almost reached by all the fleets
maximum) and that the overall process screens thousands pos- since the SM is allowed to be relatively high (30%) due to the
sible fleets, the presented application case runs in five days on a low fuel costs. The two MKRs are realised using the same num-
single core laptop. ber of ships (nS = 11) but using different CM. The Jamaica HAS
SHIPS AND OFFSHORE STRUCTURES 13

Table . Possible fleets for the complete scenario.

L B D T VS VR SM CM nS Costs
Destination (m) (m) (m) (m) (kn) (kn) (%) (–) (–) (k$)

MKR . . . . . .  .  .E+


MKR . . . . . .  .  .E+
Aruba (HAS) . . . . . .  .  .E+
Jamaica (HAS) . . . . . .  .  .E+
Curacao (HAS) . . . . . .  .  .E+

×10 4
7

6
nMR=6
Consumables costs [k$]

nMR=7
5 nMR=8
nMR=9
4 nMR=10
nMR=11
nMR=12
3
nMR=13
nMR=14
2

1
18 17 16 15 14 13 12
Reduced speed [kn]
(a) Consumables Costs.
5
×10
2

1.8 nMR=6
nMR=7
Running costs [k$]

nMR=8
1.6 nMR=9
nMR=10
nMR=11
1.4
nMR=12
nMR=13
nMR=14
1.2

1
18 17 16 15 14 13 12
Reduced speed [kn]
(b) Running Costs.

Figure . Operational costs vs. reduced speed for the MKR. Consumables costs (a) and running costs (b). (This figure is available in colour online.)

needs more ships compared to the other HAS patterns, due to logistics model able to identify the DPs of each destination and
the greater distance and higher gas request. the nominal feasible fleets in terms of number of ships, ship
capacity and ship speed. Results from this preliminary stage of
7. Conclusions the process have automatically been integrated into a second
A new SBD framework for designing fleet of ships for CNG phase dedicated to the technical design and economical assess-
trades in complex real scenarios has been proposed and applied ment of variants of the previously chosen feasible solutions. This
to the analysis of a realistic case of CNG transport. The method second part of the process is achieved by an ad hoc FSM that
follows an integrated approach that combines a preliminary includes both typical preliminary design criteria. More refined
14 T. GAGGERO ET AL.

naval architecture analyses are also included, such as the eval- CM Capacity Margin
uation of the added resistance in waves to obtain the average DP distribution pattern
statistic speed reduction and the mean consumption increase of FSM fleet synthesis model
each candidate fleet due to rough sea. The novelty of the pro- HAS Hub-and-Spoke
posed approach with respect to previously presented analyses MKR Milk-Run
and methods is in fact the full integration of the logistics into the OpEx Operational Expenditure
design process without any interruption. In addition, the logis- SM Sea Margin
tic analysis provides constraints for the further technical design SSM ship synthesis model
of the possible fleets.
The capabilities of the method have been proved by an appli-
Disclosure statement
cation example, that is the development of a new CNG trade
No potential conflict of interest was reported by the authors.
among the island of the Carribean Sea, considering a scenario
in which all the energy request of such islands is provided by the ORCID
transported natural gas. Tomaso Gaggero http://orcid.org/0000-0002-2956-5094
The results both in terms of the global transport system and Giuliano Vernengo http://orcid.org/0000-0001-8191-8742
the technical fleet designed reveal that including logistics in the Marco Parodi http://orcid.org/0000-0002-0898-8813
fleet design process gives two advantages. In fact, it allows the Enrico Rizzuto http://orcid.org/0000-0001-8112-8978
reduction of the number of solutions to be technically evaluated
since only a restricted number of combinations of (nS , VS , QS ) References
satisfy the logistic process. In addition, it provides a criterion Boulougouris E, Papanikolaou A, Østvik I, Brett PO, Konovessis D. 2012.
on the CM to be fulfilled by technically possible fleets, reducing Logistics based design as an approach to ship and business develop-
ment. Procedia Soc Behav Sci. 48:2241–2250.
further the final basin of solutions, hence giving valuable infor- Brett PO, Boulougouris E, Horgen R, Konovessis D, Oestvik I, Mermiris
mation to support the decision-making process. G, Papanikolaou A, Vassalos D. 2006. A methodology for logistics-
The proposed integrated approach might be further based ship design. Proceeding of the 9th International Marine Design
improved e.g. by using multi-objective optimisation algorithms Conference-IMDC06; Ann Arbor-Michigan.
to accelerate the selection and the convergence to the opti- CIA. 2016. The world factbook 2014–16. Government Printing Office.
https://www.cia.gov/library/publications/the-world-factbook/.
mal solution. In this perspective, operative and design related Dunlop JP, White CN. 2003. CNG transport technology is delivering on
uncertainties might be included by using multi-objective robust promises. SPE Annual Technical Conference and Exhibition; Oct 5–8;
design optimisation (MORDO) algorithms. To complete the Denver, CO. Society of Petroleum Engineers.
operation-oriented design of the fleet, a further verification of Economides MJ, Wood DA. 2009. The state of natural gas. J Nat Gas Sci
the obtained results by discrete events (DE) simulations of each Eng. 1(1):1–13.
Gaiotti M, Rizzo C, Rizzuto E, Vernengo G. 2016. Material selection for the
DP might be carried out aiming at a better understanding of the gas containment system of a compressed natural gas carrier fleet. Appl
involved time-domain phenomena such as delays, generation Ocean Res. 55:37–47.
of queues at terminal and possible storage overloads. Hsu CI, Hsieh YP. 2007. Routing, ship size, and sailing frequency decision-
making for a maritime hub-and-spoke container network. Math Com-
put Model. 45(7):899–916.
Imai A, Shintani K, Papadimitriou S. 2009. Multi-port vs. hub-and-spoke
Nomenclature port calls by containerships. Transport Res E Logist Transport Rev.
Ltrip route length (nM) 45(5):740–757.
nS number of ships per fleet Karsten CV, Pisinger D, Ropke S, Brouer BD. 2015. The time constrained
NP number of receiving ports multi-commodity network flow problem and its application to liner
qc consumption rate (MMscf/h) shipping network design. Transport Res E Logist Transport Rev.
76:122–138.
qOL offloading rate (MMscf/h) Lin DY, Tsai YY. 2014. The ship routing and freight assignment problem for
qLO loading rate (MMscf/h) daily frequency operation of maritime liner shipping. Transport Res E
QS ship nominal gas capacity (MMscf) Logist Transport Rev. 67:52–70.
QFleet fleet nominal gas capacity (MMscf) Lothe P. 2005. The Knutsen OAS shipping pressurized natural gas carrier
QR % of the required amount of energy from fossil (PNG). The Fifteenth International Offshore and Polar Engineering
Conference; Jun 19–24; Seoul, Korea. International Society of Offshore
sources at the destination port (MMscf) and Polar Engineers.
Tc terminal connection time (h) Nikolaou M. 2010. Optimizing the logistics of compressed natural gas trans-
VS ship nominal speed (kn) portation by marine vessels. J Nat Gas Sci Eng. 2(1):1–20.
VR ship reduced speed (kn) Papanikolaou A. 2010. Holistic ship design optimization. Comput Aid Des.
Tcons consumption time 42(11):1028–1044.
Papanikolaou A, Mohammed EA, Hirdaris SE. 2014. Stochastic uncertainty
TRT time for a round trip (h) modelling for ship design loads and operational guidance. Ocean Eng.
TOL time for offloading (h) 86:47–57.
TMKR time of a cyclical MKR (h) Spanos D, Papanikolaou A, Papatzanakis G. 2008. Risk-based onboard
TN travelling and connection time (h) guidance to the master for avoiding dangerous seaways. Proceeding of
nX longitudinal number of PVs the 6th International Osaka Colloquium on Seakeeping and Stability of
Ships; Mar 26–29; Osaka, Japan. p. 249–256.
nY lateral number of PVs Stenning D, Fitzpatrick J, Trebble M. 2012. Floating CNG: a simpler way
AM Adjacency Matrix to monetize offshore gas. Offshore Technology Conference; 30 Apr–3
CapEx Capital Expenditure May; Houston, TX.
SHIPS AND OFFSHORE STRUCTURES 15

Taggart R. 1980. Ship design and construction. New York (NY): Society of Table B. Most recent electricity production data.
Naval Architects and Marine Engineers.
Electricity % from Gas volume
Valsgård S, Reepmeyer O, Lothe P, Strøm NK, Mørk K. 2004. The develop-
production fossil consumption
ment of a compressed natural gas carrier. In 9th International Sympo- Destination (BkWh/yr) fuels rate (Bscm/yr)
sium on Practical Design of Ships and Other Floating Structures; Sep
12–17; Luebeck-Travemuende, Germany. Montserrat .  .
Vernengo G, Gaggero T, Rizzuto E. 2016. Simulation based design of a fleet Turks and Caicos Islands .  .
of ships under power and capacity variations. Appl Ocean Res. 61:1–15. British Virgin Islands .  .
Vernengo G, Rizzuto E. 2014. Ship synthesis model for the preliminary Anguilla .  .
design of a fleet of compressed natural gas carriers. Ocean Eng. 89:189– Dominica . . .
Antigua and Barbuda .  .
199.
Saint Vincent and Grenadines . . .
Wagner JV, van Wagensveld S. 2002. Marine transportation of compressed Saint Kitts and Nevis . . .
natural gas: a viable alternative to pipeline and LNG. In: Asia Pacific Oil Grenada . . .
and Gas Conference and Exhibition; Oct 8–10; Melbourne, Australia. Saint Lucia .  .
Society of Petroleum Engineering. Cayman Islands .  .
Wang S, Meng Q. 2012. Robust schedule design for liner shipping services. Haiti , . .
Transport Res E Logist Transport Rev. 48(6):1093–1106. Bermuda , . .
Aruba . . .
Martinique .  .
Barbados .  .
Appendices United States Virgin Islands .  .
Curacao .  .
Appendix 1. Best path for milk run distribution patterns Guadeloupe .  .
The path for each possible MKR pattern is chosen solving the Bahamas .  .
so-called Travelling Salesman Problem (TSP), that is to find the Jamaica . . .
Dominican Republic . . .
shorter path that covers a number of destinations given the dis- Puerto Rico . . .
tances between each pair of them. For the specific scenario, such
distances are included in the Adjacency Matrix AM listed in
Table A1. Values from Table A1 refer to the best solution of the
Appendix 3. Limit condition on the Capacity Margin for
TSP found using a classic genetic algorithm (GA).
the HAS and the MKR
The minimum ship capacity in a HAS solution is provided by
Appendix 2. Data on electricity production of each island Equation (3). Considering that the minimum ship capacity QS
Electricity production data of each of the island of the Carribean can be improved by a CM, resulting in a new ship capacity Q1S =
Sea included in the analysed scenario are listed in Table B1 QS · CM and assuming:
(CIA 2016). Each data corresponds to the most updated avail-
able value at the Summer of 2016. In the proposed test case, it
has been assumed that all the energy needs of these island will qOL
c1 = qOL , (C1)
be completely replaced by CNG. nS − 1 − qL

Table A. Adjacency Matrix of distances (nm) between Caribbean Islands: -Trinidad, -Montserrat, -Turks, -Brit. Virgin, -Anguilla, -Dominica, -Antigua, -S. Vincent, -
S. Kitts, -Grenada, -S. Lucia, -Cayman, -Haiti, -Bermuda, -Aruba, -Martinique, -Barbados, -US Virgin, -Curacçao, -Guadeloupe, -Bahamas, -Jamaica,
-Dominican Rep., -Puerto Rico.

–                        

                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
                        
16 T. GAGGERO ET AL.

1.12 CM Limit Curve ×10 6


FSM results 1.4
CM=1
1.1 CM=1.025
CM=1.05
1.35
CM=1.075
1.08
CM=1.1

Fleet Cost [k$]


1.06 1.3
CM

1.04
1.25
1.02

1.2
1

0.98
1.15
11.2 11.3 11.4 11.5 11.6 11.7 11.8 11.9 12 12 11.9 11.8 11.7 11.6 11.5 11.4 11.3 11.2
VR Reduced speed [kn]

Figure C. Example of solutions for the Aruba HAS in terms of CM against the ship Figure C. Example of solutions for the Aruba HAS in terms of fleet costs against
reduced speed VR . The curve corresponding to the limit condition of Equation (C) reduced speed VR . Solutions discarded imposing the limit condition of Equation (C)
is shown too by a red line. (This figure is available in colour online.) are highlighted by blue dots. (This figure is available in colour online.)

with the results of the preliminary logistics model. As evident


from this example, none of the fleets designed with CM = 1 ful-
c2 = 4Tc , (C2) fil the requirement of Equation (C5).
c3 = 2Ltrip . (C3) Following the same approach, considering the minimum
capacity of a ship in a MKR distribution pattern obtained simply
Equation (3) can be rewritten in the form: dividing Equation (8) by nS and assuming:

qc qc
CM ≥
c1 c 2
+
c 1 c3
. (C4) c4 = n − − , (C6)
QS QSVS qOL qLO
c5 = 2Tc qc (NP + 1), (C7)
Being the terms α = and β = c1 c2
QS
c1 c3
and considering that each
QS c6 = qc Ltrip . (C8)
solution designed by the FSM comes with a reduced speed VR
due to the rough sea effects, the limit condition for the CM in a Equation (8) can be rewritten as
HAS solution is then given by Equation (C5):
c5 c6
CM ≥ + . (C9)
β QS c4 QS c4VS
CM ≥ α + . (C5)
VR
Being the terms γ = (QcS5c4 ) and φ = (QcS6c4 ) and considering again
An example of the influence of this condition is given in that each solution designed by the FSM comes with a reduced
Figures C1 and C2 for the Aruba HAS. For the given scenario, speed VR due to the rough sea effects, the limit condition for the
all the represented solutions are technically possible for the FSM CM in an MKR solution is then given by Equation (C10):
but only those above the limit line (the red curve) are consistent
φ
CM ≥ γ + . (C10)
VR

Potrebbero piacerti anche