Ocean Engineering 152 (2018) 316–333

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Arrangement method of a naval surface ship considering stability,

operability, and survivability
Sun-Kyung Jung a, Myung-Il Roh b, *, Ki-Su Kim a
a
Department of Naval Architecture and Ocean Engineering, Seoul National University, 1, Gwanak-ro, Gwanak-gu, Seoul, 08826, Republic of Korea
b
Department of Naval Architecture and Ocean Engineering & Research Institute of Marine Systems Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu,
Seoul, 08826, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: The preliminary design stage for a naval surface ship requires several performance parameters to be considered,
Arrangement design including stability, operability, and survivability. These parameters have to be considered for many compartments
Naval surface ship constrained by a limited space, so this problem is complex and with many possible design alternatives. However,
Stability the arrangement during the actual design of a ship depends heavily on having experienced designers and parent
Operability ships. There is no mechanism to quantitatively evaluate the arrangement design. For these reasons, the probability
Survivability
of changes in the design increases as the design progresses.
Multistage optimization
This study proposes a method to arrange a naval surface ship while considering the stability, operability, and
survivability, and a prototype program based on the proposed method is herein developed. First, a template model
is defined consisting of a data structure to represent the arrangement information of the ship. Second, methods are
studies to quantitatively evaluate the stability, operability, and survivability of the ship, by referring existing
studies. Third, two-stage optimization problems for the arrangement design are mathematically formulated,
selecting design variables, objective functions, and constraints. Finally, a user interface is developed to input the
given data and to output the optimization result for the arrangement design. The applicability of the proposed
method is verified by implementing an example of a 7000-ton class missile destroyer. The result shows that the
proposed method can be effective in producing better alternatives for the arrangement design during the pre-
liminary design stage.

1. Introduction
A naval surface ship is designed to participate in several missions in
battle, so it must be suitable for exposure to human-made hostility. This is
the fundamental reason for which such ships must have different design
concept from conventional ships. In addition, numerous crew members
must live on the ship for extended periods, and thus their living condi-
tions must also be considered in the arrangement design. In particular,
the arrangement design is one of the most important parts because it
defines the design philosophy of the ship. As such, many compartments,
including tanks, rooms, and equipment, are placed in limited space on
the ship. There are many alternatives for the arrangement design, and our
hope is to select a better alternative (if possible, the best one) in less time.
The performance parameters have to be considered during the
arrangement design, including the stability, operability, and survivability
of the ship. First, the stability refers to the ability to restore the ship when
it is inclined for some reason. Although stability is a basic condition for
any ship, not just naval surface ships, it is also necessary to consider
stability even after damage due to an attack, unlike with conventional
ships. The stability of a naval surface ship is considered for when the ship
is intact or damaged. Second, the operability is the ability for effective
flow of the crew and supplies throughout the ship. Considering the
characteristics of the naval surface ship, where crew and weapons share
space, the operability has to be considered for efficient logistic movement
within the ship. Finally, the survivability is the ability for the ship to
survive under various types of threat. The survivability can be regarded
as the most important consideration for a naval surface ship when in
battle, and it is also an important part of the design of a naval surface
ship. Many things are considered when analyzing the survivability of a
ship, which can be measured through various interpretations during the

* Corresponding author. Department of Naval Architecture and Ocean Engineering & Research Institute of Marine Systems Engineering, Seoul National University, 1, Gwanak-ro,
Gwanak-gu, Seoul, 08826, Republic of Korea.
E-mail addresses: snoopyyha@snu.ac.kr (S.-K. Jung), miroh@snu.ac.kr (M.-I. Roh), kisu2511@snu.ac.kr (K.-S. Kim).

https://doi.org/10.1016/j.oceaneng.2018.01.058
Received 22 August 2017; Received in revised form 29 November 2017; Accepted 13 January 2018
Available online 23 February 2018
0029-8018/© 2018 Elsevier Ltd. All rights reserved.
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 1
Summary of related studies and comparison with this study.

Studies Target Compartment arrangement Consideration of Optimization

Hull (Bulkhead) Superstructure (Room) Stability Operability Survivability

Byun (1998) Commercial ship ✓ – ✓ – – –

Shin et al. (2002) Machinery space ✓ – – ✓ – –
(Equipment)
Lee et al. (2003, 2005) Naval surface ship – ✓ – ✓ – ✓
Boulougouris and Papanikolaou (2004) Naval surface ship ✓ – ✓ ✓ ✓ ✓
Helvacioglu and Insel (2005) Commercial ship ✓ – – – – –
Andrews (2006) Naval surface ship ✓ ✓ ✓ ✓ ✓ –
(User preference)
Nick (2008) Naval surface ship – ✓ – ✓ – ✓
Oers et al. (2008) Naval surface ship ✓ (2D) ✓ (2D) – ✓ ✓ ✓
Parsons et al. (2008) Naval surface ship ✓ ✓
Roh et al. (2009) Naval surface ship ✓ – ✓ ✓ – ✓
Chung et al. (2011) Submarine ✓ – – ✓ – ✓
Shin (2013) Naval surface ship – ✓ – – ✓ –
Ju et al. (2014) Naval surface ship – ✓ – ✓ ✓ ✓
Kim et al. (2015) Submarine ✓ – – ✓ – –
Kim and Roh (2016) Submarine ✓ – ✓ ✓ – ✓
Jung (2016) Naval surface ship – ✓ – – ✓ ✓
This study Naval surface ship ✓ ✓ ✓ ✓ ✓ ✓

actual design of the ship. The arrangement design in the preliminary
design stage is important because the earlier the design stage is, the less
the cost associated with the design, and at the same time, the higher the
flexibility of the design. Therefore, it is important to carefully consider
the performance mentioned above in the preliminary design stage.
As mentioned above, the naval surface ship has a number of com-
partments that must fit in limited space, and there are many alternatives
for its arrangement. Thus, an optimization process may find the better or
best arrangement for the naval surface ship while considering many
criteria, including the stability, operability, and survivability. Of course,
the criteria mentioned above are necessary for merchant ships for a safe
return to port, as well. The problem mentioned above is difficult to solve,
but an optimization technique can efficiently solve it within the allow-
able time. Therefore, this study involves a mathematical formulation for
the optimization problem of finding a better arrangement. However, it is
very difficult to formulate an optimization problem while considering all
variables in the arrangement design, so the optimization process is sub-
divided into two stages: bulkhead arrangement in the hull and room
arrangement in the superstructure.
2. Related studies
Several researchers have proposed methods to solve arrangement
problems of ships, and in this section, we summarize past studies related
to arrangement design in naval architecture. Byun (1998) proposed a
rule-based expert system incorporating a knowledge base to support the
compartment design of a commercial ship. Shin et al. (2002) proposed an
expert system to arrange the machinery on a ship. They proposed an
expert system to evaluate the feasibility of design alternatives for ma-
chinery arrangement. Lee et al. (2003, 2005) used an improved genetic
algorithm to arrange rooms and passages in the superstructure of a naval
surface ship taking into account its operability, including the flow of
crews and supplies. Boulougouris and Papanikolaou (2004) proposed a
method to arrange compartments in the hull of a naval surface ship and
also proposed methods to assess the stability and survivability. They then
formulated an optimization problem for the compartment arrangement
and solved it using a multi-objective genetic algorithm (MOGA). Helva-
cioglu and Insel (2005) proposed a multistage expert system to arrange
the compartments on a container ship.
Andrews (2006) proposed the building block method to perform the
initial arrangement design of a naval surface ship. The building block
method allows the arrangement design according to various design re-
quirements and designer's preferences, and it also enables naval archi-
tectural calculations. Nick (2008) proposed a method to generate,
evaluate, and optimize the compartment arrangement of a naval surface
ship. She defined the allocation as the assignment of space (compart-
ment) to a deck of a ship, and the arrangement is used to determine the
topology and geometry of the space for the assignment of each deck. Oers
et al. (2008) proposed an optimization method for the arrangement
design of a naval surface ship. They determined a Pareto-optimal set from
various objective functions. Parsons et al. (2008) conducted zone-deck
optimization and detailed room arrangement in the zone-deck through
fuzzy optimization and allocation optimization. That is, they performed
the arrangement design with two steps: allocation and arrangement. In
their study, the relationship between compartments and the re-
quirements of designers were reflected. Roh et al. (2009) proposed a
method to arrange bulkheads in the hull of a naval surface ship. They
formulated an optimization problem for the bulkhead arrangement of
which objective functions are to maximize the space for the armament
and the structural strength while satisfying stability requirements. Chung
et al. (2011) proposed a rule-based expert system to optimize the com-
partments in the pressure hull of a submarine. Shin (2013) proposed a
method to arrange rooms in the superstructure of a naval surface ship by
considering its survivability. He tried to quantify the survivability by
using FMEA (Failure Mode and Effective Analysis) and FTA (Fault Tree
Analysis). Ju et al. (2014) proposed a method to arrange rooms in the
superstructure of a naval surface ship. He used the SLP (Systematic
Layout Planning) method to solve an optimization problem for the room
arrangement of which objective functions are to optimize the surviv-
ability and operability. Kim et al. (2015) proposed an expert system to
arrange the compartments and equipment in the pressure hull of a sub-
marine. The proposed expert system was based on the arrangement
evaluation model (AEM) to evaluate the feasibility of the design alter-
natives and the arrangement template model (ATM) to store the
arrangement information. Kim and Roh (2016) proposed an arrangement
method of a submarine based on the expert system and the optimization
technique. In their study, the arrangement design was mathematically
formulated as a three-stage optimization problem for compartments,
sub-compartments, and equipment. Jung (2016) proposed a method to
arrange rooms in the superstructure of a naval surface ship and also
proposed a method to evaluate the integrated survivability considering
the vulnerability, susceptibility, and recoverability.
As seen above, the most relevant studies have focused separately on
the compartment (bulkheads or rooms) arrangement in the hull or in the
superstructure, but not both. Furthermore, not all performance, such as
stability, operability, and survivability, were considered, even though it
is important to consider them simultaneously in the design of naval
surface ships. Thus, this study proposes a method to optimally arrange

317
S.-K. Jung et al.
Fig. 1. Configuration of the proposed arrangement method of a naval surface ship.
Fig. 2. GZ and HA curves for the intact stability criteria by the US Navy.
bulkheads and rooms considering the major aspects of the performance
as a naval surface ship. Table 1 shows a summary of related studies and a
comparison of those studies with this study.
The remainder of this paper is organized as follows. Section 3 de-
scribes the method proposed in this study for the arrangement design of a
naval surface ship. In Section 4, the proposed arrangement method is
applied to a bulkhead arrangement and a room arrangement, and the
results of an actual application are discussed. The final section presents
the overall results of this study and briefly discusses the remaining work
left for future study.
3. Arrangement design method of a naval surface ship
This study proposes a method for the arrangement design of a naval
surface ship. Fig. 1 shows the configuration of the proposed method. As
shown in this figure, the proposed method consists mainly of six com-
ponents: the template model, stability module, operability module,
318
Ocean Engineering 152 (2018) 316–333
survivability module, optimization module, and user interface.
The arrangement design requires a data structure that can store
arrangement information. In this study, the template model (Fig. 1 (1)) is
used as the data structure for the arrangement design of the naval surface
ship. It was hierarchically defined to store various pieces of information
related to the bulkheads, decks, compartments, tanks, and rooms. Also,
the template model is used to interface various information among the
other modules (Fig. 1 (2) to (5)) and the user interface (Fig. 1 (6)).
To evaluate various performance parameters, including the stability,
operability, and survivability of the given alternative for the arrangement
design, the stability module (Fig. 1 (2)), the operability module (Fig. 1
(3)), and the survivability module (Fig. 1 (4)) were developed. In addi-
tion, bending moment was calculated separately. These modules give us
quantitative values for each performance parameter using a reasonable
method, and the values are used in the optimization module as objective
functions and constraints for the optimization.
In the field of engineering, optimization techniques have gained
widespread use to obtain an optimal design according to specific re-
quirements. In this study, an optimization technique is used to derive the
arrangement design of a naval surface ship. Thus, the optimization
module (Fig. 1 (5)) was developed, and it includes the mathematical
formulation of the optimization problems and an optimization algorithm
to solve the formulated problems.
The user interface is also needed for the designer to easily utilize the
components that are mentioned above. In this study, the user interface
(Fig. 1 (6)) was developed as a tool for the input and output of the
arrangement design, including a function of 3D visualization.
3.1. Stability module for the arrangement design of a naval surface ship
The stability module is used to assess the stability of a naval surface
ship, under intact and damaged conditions. Such assessment can be done
by verifying some stability criteria.
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 2 3.1.2. Evaluation of the damage stability

Criteria for the intact stability by the US Navy. The damage stability of a naval surface ship can be roughly checked
No Criteria Required value using a deterministic approach. The basis to determine the extent of
1 GZ0/GZmax 0.6 flooding is the length of damage to the shell at any point along the ship's
2 Ratio between capsizing and restoring energy (A2/A1) 1.4 length resulting from an attack with a weapon or a collision. In particular,
3 Equilibrium heel angle (ϕ0) 15 deg the combatant-type naval surface ships of over 90 m in length should
meet criteria for adequate subdivision to resist underwater damage
(Sarchin and Goldberg, 1962). That is, they should be capable of with-
Table 3 standing flooding from a shell opening equal to 12.5% of the ship's length
Criteria for the damage stability by US Navy. (longitudinal extent) at any point fore and aft. The maximum transverse
No Criteria Required value extent is assumed to be that caused by penetrating damage to, but not
including, any center bulkhead, and the maximum vertical extent is
1 Equilibrium heel angle (ϕ0) 15 deg
2 Ratio between capsizing and restoring energy (A2/A1) 1.4 assumed for all decks to be opened because of the effect of the resulting
3 GZmax - HA 0.075 m high flooding, free surface, and possible unsymmetrical flooding. With
4 Area A1 0.02 m⋅rad this assumption for the extent of the damage, satisfying the damage
5 Distance from waterline to margin line (Dist) 0.0 m
stability can be checked with the criteria summarized in Table 3.
Referring to Fig. 3 and Table 3, the equilibrium heel angle (ϕ0) should
be less than 15 deg. The ratio of area A2 to A1 should be greater than 1.4,
and the difference between GZmax and HA at the angle corresponding to
GZmax should be greater than 0.075 m. The area A1 should be greater than
0.02 m⋅rad. Finally, the margin line should be higher than the waterline
in the damaged state (Republic of Korea Navy, 2014; Kim et al., 2017a).

3.2. Operability module for the arrangement design of a naval surface ship

The operability can be referred to as the ability to represent the

effectiveness of the flow of crews and supplies in the ship. In this study,
the operability was represented with the adjacency index (AI) that is
calculated using the adjacency coefficient (qij) and the distance (dij) be-
Fig. 3. GZ and HA curves for the damage stability criteria by US Navy. tween the compartments i and j, as shown in Eq. (1) (Kim et al., 2017b).

3.1.1. Evaluation of intact stability X

To verify the criteria for the intact stability, a GZ (righting arm) curve
and an HA (heeling arm) curve should be calculated, as shown in Fig. 2.
To calculate the GZ curve, the buoyancy and weight changes should be
considered for different heel angles. Thus, if the positions of the bulk-
heads associated with liquid tanks change, the GZ curve should be
calculated considering the free surface effect of the liquid tanks (Moore
and Paulling, 2010). With these curves, the adherence to the conditions
for intact stability can be checked, as summarized in Table 2 (Brown and
Deybach, 1998).
Referring to Fig. 2 and Table 2, the ratio of the righting arm at the
intersection of the GZ and HA curves (GZ0) to the maximum righting arm
(GZmax) should be less than 0.6. The ratio between capsizing and
restoring energy, that is, the ratio of area A2 to A1 should be greater than
1.4. Finally, the equilibrium heel angle (ϕ0) should be less than 15 deg for
some severe cases, such as lifting heavy weights over the side, crowding
of crews to one side, high-speed turning, and so on.
N c 1 X
Nc
AI ¼ qij ⋅dij (1)
i¼1 j¼iþ1
In Eq. (1), the symbol Nc is the number of compartments.
3.2.1. Adjacency index between compartments
The adjacency coefficient is a constant that quantitatively represents
the degree of closeness between compartments, and it can be calculated
from the affinities and antagonism, as shown in Eq. (2) (Kim et al.,
2017b).
qij ¼ Affinityij  Antagonismij (2)
In Eq. (2), the affinity is the characteristic that makes it particularly
advantageous to locate one compartment close to another specific
compartment, and it is related to the movement of crews and supplies
between compartments. A frequent movement of crews and supplies is
anticipated between specific compartments; then it becomes advanta-
geous to locate these compartments close to each other. The antagonism
represents the characteristics that preclude a compartment from being

Fig. 4. Distance calculation between the compart-

ments considering the passages.

319
S.-K. Jung et al.
separately and safely located near another specific compartment. As a
consequence, the adjacency coefficient between each compartment can
be calculated by subtracting the antagonism from the affinity. The ad-
jacency coefficients (qij) between the compartments can then be deter-
mined by a designer and are given as one of the input data for the
arrangement design.
3.2.2. Distance calculation between compartments
There are many rooms, including passages, in the superstructure.
Thus, when we calculate the distance (dij) between the compartments
(i.e., the room) i and j, the passages between them should also be
considered. In this study, Dijkstra's algorithm from graph theory was
used to determine an accurate distance (Lee et al., 2003, 2005).
For example, to find the shortest path between compartments 3 and
10, and its distance, all possible relationships among compartments 3,
10, and the passages can be represented in the adjacency graph, as shown
in Fig. 4 (b). In the adjacency graph of Fig. 4 (b), each node represents the
compartment (3 and 10) or the passage (a, b, c, and d). Each edge rep-
resents the distance between the compartment and horizontal passage or
between the horizontal and vertical passages. When using Dijkstra's al-
gorithm, the length of each edge must be calculated to determine the
shortest path, and its distance, between compartments i and j. To do this,
two starting points for the distance calculation called “base points”
should first be defined. In this study, a centroid (center of area) of the
compartment is used as the base point for the compartment. However, for
the passage, it is difficult to define the base point. If the centroid of the
passage is used as the base point, the redundant distances may be
included in the resulting distance between the compartments.
To avoid making this error, the base points for the vertical and hor-
izontal passages are defined as follows. For the vertical passage (b and c),
the centroid of the passage is used as the base point. For the horizontal
passage (a and d), the y- and x-coordinates of the base point are assumed
as the vertical centroid of the horizontal passage (for the y coordinate)
and (xf þ xp)/2 (for the x-coordinate). Here, xf and xp represent the
horizontal centroids of the compartment and the vertical passage,
respectively. For example, the x-coordinate (xpa1) of the base point for the
horizontal passage (a1) is assumed as the horizontal center point be-
tween compartment 3 and the vertical passage (b). The y-coordinate (ypa1)
is assumed as the vertical center point of the horizontal passage (a1), as
shown in Fig. 4 (a). An example defining the base points for the hori-
zontal passages is shown in Fig. 4 (a), and the corresponding adjacency
graph is shown in Fig. 4 (b).
3.3. Survivability module for the arrangement design of a naval surface
ship
According to Said (1995), the survivability is defined as the capability
of a ship and its onboard systems to avoid and withstand a weapons ef-
fects environment without sustaining impairment of their ability to
accomplish designated missions. The survivability has three components:
320
Ocean Engineering 152 (2018) 316–333
Fig. 5. Calculation of the explosion pressure of the
bulkhead due to an explosive.
vulnerability, susceptibility, and recoverability. The vulnerability refers
to the inability of a ship to withstand damage from one or more attacks,
its vincibility, and to its liability to serious damage or loss when attacked
by weapons. The susceptibility refers to the inability of a ship to avoid
being damaged in the pursuit of its mission and to its probability of being
attacked. Lastly, the recoverability refers to the ability of a ship and its
crew to prevent loss and restore mission-essential functions from an
attack due to one or more weapons. Among these, the vulnerability can
be improved by adjusting the positions of the bulkheads and decks and
the locations of rooms. Therefore, this study handles the vulnerability as
a major component that represents the survivability of the ship.
Regarding the arrangement design of a naval surface ship, two vulnera-
bility scenarios for the bulkheads in the hull and for rooms in the su-
perstructure are defined and evaluated as follows.
3.3.1. Bulkhead damage vulnerability
If an explosion due to an attack occurs within the compartment of a
naval surface ship, the pressure of the explosion will act on adjacent
bulkheads and decks, as shown in Fig. 5. At this time, if the pressure due
to the explosion can be minimized, we can say that the vulnerability of
the ship may be reduced. In this sense, minimizing all explosion pressure
of the bulkheads induced by all explosives may be a criterion to select a
better alternative in the arrangement design in terms of the vulnerability.
Thus, the bulkhead damage vulnerability (VBD) can be represented using
Eq. (3).
X
N B&D
VBD ¼ Pir (3)
i¼1
In Eq. (3), the symbol Pir is the explosion pressure of the ith bulkhead
or deck due to an explosive, and the symbol NB&D is the number of
bulkheads and decks in the hull. The explosion pressure (Pir) of the
bulkhead or deck due to an explosive can be calculated using Eq. (4) (US
Department of the Army, 1990; Choi et al., 2015). This equation is an
empirical fit equation for the reflected pressure according to the distance
from the explosive point to the bulkhead or deck.
(  2 )
ðγ þ 1Þ⋅ PiSnd
Pir ¼ 2⋅P0 ⋅ PiSnd þ (4)
ðγ  1Þ⋅PiSnd þ 2γ
In Eq. (4), the symbol P0 is the ambient air pressure in Pa (¼
101,300 Pa at sea level), the symbol PSnd is the non-dimensional side-on
overpressure in Pa, defined as Eq. (5), and the symbol γ is the specific
heat ratio of P0 and PSnd, taken as 1.4.
( 1=3 )1:95
P0
PiSnd ¼ 0:55⋅ Ri ⋅ (5)
EW ⋅W
In Eq. (5), the symbol Ri is the distance from the center of the ex-
plosion to the ith bulkhead or deck in m, the symbol EW is the explosive
S.-K. Jung et al.
energy per unit mass in m2/s2, and the symbol W is the charge mass of an
explosive in kg.
In Eq. (4), Pir is a function of the distance of bulkheads or decks from
the explosive point. Therefore, to calculate the vulnerability of the ship,
the positions of explosives corresponding to threats have to be defined.
The types of threats also have to be selected. In this study, anti-ship
missile (ASMs), torpedoes, and mines were considered as major threats
to the ship. For each threat, the longitudinal damage distribution was
assumed by referring to existing studies (Harmsen and Krikke, 2000;
Boulougouris and Papanikolaou, 2004). We assumed that ASMs follow a
piecewise linear distribution along the ship's length, torpedoes follow a
normal distribution, and mines follow a linear distribution, as shown in
Fig. 6.
As shown in Fig. 6, the distribution of the ASMs has its peak point at
midship because the midship part has the largest area being detected, and
the remaining part shows a linear change. Due to the characteristics of
torpedos, which target a ship by sensing the noise radiated underwater
that is usually generated by an engine and propeller, the midpoint of the
normal distribution slightly moved to the stern of the ship. In the case of
mines, the probability of damage of the stern is zero, and the probability
of damage to the bow side behaves linearly, as they will explode by the
pressure over which the ship passes. In this study, for all threats, the
explosive energy per unit mass (EW) was assumed to be 4.184⋅109 m2/s2,
and the charge mass (W) of the ASMs, torpedoes, and mines was assumed
to be 161 kg, 300 kg, and 113.4 kg, respectively, by referring to existing
studies (Harmsen and Krikke, 2000; Boulougouris and Papanikolaou,
2004).
321
Ocean Engineering 152 (2018) 316–333
Fig. 6. Longitudinal damage distribution of ASMs,
torpedoes, and mines.
Fig. 7. Example of the probability of damage for
torpedoes.
3.3.2. Room damage vulnerability
Similarly, the concept of the vulnerability can be applied to rooms in
the superstructure. In this study, the room damage vulnerability is
defined as a summation of the multiplication of the probability being
damaged and the importance factor of each room, as shown in Eq. (6).
X
NR
VRD ¼ pi ⋅ei (6)
i¼1
In Eq. (6), the symbol pi and ei are the probability of damage and the
importance factor of the ith room, respectively, and the symbol NR is the
number of rooms in the superstructure.
(1) Probability of damage of the room
A naval ship has a probability of being attacked along her length,
breadth, and depth. If a certain room is placed in a certain position of the
superstructure, the room has its own probability of damage according to
its x-, y-, and z-positions, as shown in Fig. 7.
In this study, the probability of damage of the ith room is defined as
the multiplication of the probability of damage length (pDL) and the
probability of impact (pIP), as shown in Eq. (7) (Przemieniecki, 1994;
Boulougouris and Papanikolaou, 2004).
pi ¼ piDL ⋅piIP (7)
The probability of the damage length (pDL) is again defined as Eq. (8).
That is, pDL means the fraction of the target (e.g., room) that is assumed to
have been damaged within a certain radius from the explosive point of an
explosive.
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

a
piDL ¼ ∫ 0i Damðai Þ
"
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
1 ln2 ai RSS ⋅RSK (8)
a
¼ ∫ 0i pffiffiffiffiffi n  ⋅exp   2
1 1
2π ⋅ pffiffiffi lnðRSS =RSK Þ ⋅ai 2⋅ pffiffiffi lnðRSS =RSK Þ
2 2⋅1:45222 2 2⋅1:45222

 
5
Impy ðyÞ ¼ 2 ðy  0:5Þ2 þ for ASM; torpedoes; and mines (13)
12
In Eq. (8), the symbol ai is the longitudinal, transverse, or vertical
Finally, function Impz(z) is the impact density function along the
distance between the centroid of the ith room and the explosive point
vertical direction for each threat. In this study, we assume that this
which is normalized by the ship's length, breadth, or depth, whichever is
probability for all threats follows a linear distribution along the ship's
lesser. RSK is the sure kill radius that is assumed in this study as 0.02L (L:
depth, including the superstructure, as shown in Eqs. (14) and (15).
ship's length), and RSS is the sure save radius that is assumed as 0.15L by
referring to existing studies (Boulougouris and Papanikolaou, 2004). Impz ðzÞ ¼ 2 for ASM (14)
The probability of impact (pIP) is defined as Eq. (9).

xi ai =2 yi ai =2 zi ai =2 Impz ðzÞ ¼ 2z þ 2 for torpedoes and mines (15)
piIP ¼ ∫ x2i þai =2 Impx ðxÞdx⋅∫ y2i þai =2 Impy ðyÞdy⋅∫ z2i þai =2 Impz ðzÞdz (9)
1 1 1
When calculating the room damage vulnerability, the longitudinal
In Eq. (9), the coordinates xi1 and xi2 are the x-coordinates aft and damage distribution for each threat was assumed as the same as that of
forward, which are normalized by the ship's length, of the ith room, the bulkhead damage vulnerability in Fig. 6.
respectively. The coordinates yi1 and yi2 are y-coordinates of the port and
starboard, which are normalized by the ship's breadth, of the ith room, (2) Importance factor of the room
respectively. The coordinates zi1 and zi2 are the z-coordinates of the bot-
tom and top of the ith room, respectively, and these are normalized by the If a specific room, including equipment is damaged, such equipment
ship's depth, including the superstructure. can also have damage. In this sense, if a room has important equipment,
Function Impx(x) is the impact density function along the longitudinal the room should be more carefully considered during the arrangement
direction for each threat. Considering that the ASMs follow a piecewise design. Considering this fact, the importance factor (ei) of each room can
linear distribution along the ship's length, the torpedoes follow a normal be defined by a summation of the multiplication of the importance co-
distribution, and the mines follow a linear distribution, Impx(x) for each efficient and the number of each piece of equipment in the room, as
threat are assumed to follow Eqs. (10)–(12). shown in Eq. (16).

4x if x  0:5 X
NR i
Impx ðxÞ ¼ for ASM (10) ei ¼ Ie ik ⋅Ne ik
4x þ 4 if x > 0:5 (16)
k¼1
 
1 1 In Eq. (16), the symbol NiR is the number of all pieces of equipment in
Impx ðxÞ ¼ pffiffiffiffiffi exp  2 ðx  0:4Þ2 for torpedoes (11)
2π ⋅σ 2σ the ith room. The symbols Ike and Nik e are the importance factor (between
0 and 1) and the number of the kth equipment in the ith room, respec-
In Eq. (11), the symbol σ is standard deviation of normal distribution. tively. The importance factor (Iik
e ) for all equipment can be determined by
 a designer and are given as input data for the arrangement design.
0 if x  0:5
Impx ðxÞ ¼ for mines (12)
8x þ 4 if x > 0:5
3.4. Optimization module for the arrangement design of a naval surface
Similarly, function Impy(y) is the impact density function along the ship
transverse direction for each threat. In this study, we assume that this
probability for all threats follows parabolic distribution along the ship's This section describes an optimization method for the arrangement of
breadth, as shown in Eq. (13). a naval surface ship. As mentioned earlier, the arrangement method for
the naval surface ship was formulated in two stages in this study. The first
stage is the bulkhead arrangement where the positions of the bulkheads
and decks in the hull are determined by minimizing the maximum
bending moment at the intact state, the space for liquid tanks, and the
bulkhead damage vulnerability due to the attack. The second stage is the
room arrangement where the sequence of rooms and the positions of
passages in the superstructure are determined by minimizing the adja-
cency index between the rooms and the room damage vulnerability due
to the attack. Since the changes of the second stage such as weight dis-
tribution are unknown in the first stage, the constraints related to sta-
bility in the first stage are confirmed again after the optimization of the
second stage. In other words, we check the stability criteria again after
two optimization stages whether the derived arrangement is feasible or
not. The configuration for the optimization procedure is shown in Fig. 8.
Fig. 8. Configuration of the two stages of an optimization problem for the For the optimization at each stage, the mathematical formulation of
arrangement design of a naval surface ship. the optimization problems is described below.

322
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 4 Table 5
Input data for the bulkhead arrangement. Design variables for the optimization of the bulkhead arrangement.

Input data Notation Design Description

variables
Hull form and compartment model –
Number of bulkheads and decks NB&D xi x-coordinate of the bulkhead i (from the AP, positive forward)
Locations of compartments in the hull, including – yi y-coordinate of the bulkhead i (from the center line, positive
liquid tanks towards the port side)
Weight and weight distribution at full load condition – zi z-coordinate of the deck i (from baseline, positive upward)
and minimum operating condition
Required volumes for liquid tanks (FOTs, FWTs, Vmin max min max
FOT, VFOT , VFWT, VFWT,
WBTs, and LOTs) Vmin
WBT , V max
WBT , Vmin
LOT , V max
LOT the bulkhead arrangement were considered as the objective function and
Longitudinal damage distribution of all threats – the constraints, respectively. To prevent the space for liquid tanks from
Explosive energy per unit mass for all threats EW
Charge mass of ASMs, torpedoes, and mines WA, WT, WM
becoming too small during the optimization, the required volume for the
Minimum length for each of engine rooms lmin
ER
liquid tanks (FOTs, FWTs, WBTs, and LOTs) should also be prepared. To
Minimum deck height hmin
deck evaluate the bulkhead damage vulnerability due to the attack in Section
3.3.1, the longitudinal damage distribution, the explosive energy per unit
mass, and the charge mass of the ASMs, torpedoes, and mines are
3.4.1. Formulation for the bulkhead arrangement
required. Finally, the minimum length for each of the engine rooms and
The bulkhead arrangement in the hull is carried out in the first stage,
the minimum deck height should be input to prevent those values from
and several things need to be considered for this. Regarding the strength,
becoming too small during the optimization. These input data for the
combat performance, and survivability of a naval surface ship, the po-
bulkhead arrangement can be summarized in Table 4.
sitions of the bulkheads and decks in the hull are important during the
bulkhead arrangement. To ensure the structural strength, the maximum
(2) Design variables
bending moment at the intact state should be minimized. In addition, to
ensure the combat performance, the space for the armament should be
In this stage, the design variables are the positions of the bulkheads
maximized, that is, the auxiliary space for the liquid tanks such as fuel oil
and decks in the hull. Fig. 9 and Table 5 show the design variables for the
tanks (FOTs), fresh water tanks (FWTs), water ballast tanks (WBTs), and
optimization of the bulkhead arrangement.
lubrication oil tanks (LOTs) should be minimized. Finally, to ensure a
sufficient survivability of the ship, the bulkhead damage vulnerability
(3) Objective functions
due to an attack should be minimized. For these reasons, the process for
the bulkhead arrangement was formulated as an optimization problem in
As previously mentioned, it is very important to ensure the structural
this study.
strength, combat performance, and survivability of a ship in the bulkhead
arrangement. Therefore, we introduced three objective functions to
(1) Input data
ensure these are obtained in the first stage of the optimization.
The first objective function of this stage minimizes the maximum
When performing a bulkhead arrangement, a hull form and
bending moment at the intact state. At this time, two loading conditions,
compartment model of the naval surface ship is required. Such infor-
such as the full load condition and minimum operating condition, were
mation includes the number of bulkheads and decks, and the locations of
considered in this study. For each loading condition, the maximum
the compartments, including the liquid tanks of the ship. Also, the in-
bending moment can be obtained by calculating the buoyancy curve, the
formation of the weight and the weight distribution at the loading con-
weight curve, the load curve, the shear force curve, and the bending
ditions, such as minimum operating condition and full load condition, are
moment curve in the longitudinal direction of the ship. The bending
also needed to assess the stability requirement. In this study, only two
moment can be obtained from the weight curve and the buoyancy curve
loading conditions that have a high possibility to cause a severe state of a
of a ship under the given loading condition. Here, the weight curve
ship were considered. Here, the minimum operating condition means
represents weight distribution of the ship in the longitudinal direction.
that a ship carries the minimum amount of cargo required for operation
Thus, if the positions of bulkheads change, the weight curve changes. The
and it can cause a severe state in terms of the stability because the draft is
load curve is obtained by summing up two curves. If the load curve is
relatively low. The full load condition means that a ship carries all cargo
integrated along the longitudinal direction, the shear force curve can be
fully and it can cause a large amount of bending moment. Of course, if we
obtained. If the shear force curve is integrated again along the longitu-
consider more loading conditions, the optimal design will be more
dinal direction, the bending moment curve can be obtained. This
robust. The bending moment and the stability during the optimization of

Fig. 9. Design variables for the optimization of the

bulkhead arrangement.

323
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 6
Objective functions for the optimization of the bulkhead arrangement.

Objective functions Type Mathematical representation

Maximum bending moment at the intact state Minimize F1 ¼ maxðBM1 ; BM2 Þ

(1: full load condition, 2: minimum operating condition)
Auxiliary space for liquid tanks Minimize F2 ¼ VFOT þ VFWT þ VWBT þ VLOT
P B&D i
Bulkhead damage vulnerability due to the attack Minimize F3 ¼ VBD ¼ Ni¼1 Pr

objective function is intended to minimize the maximum value of the g13 ¼ VWBT
min
 VWBT  0 (29)
maximum bending moments for the two loading conditions. The second
objective function minimizes the auxiliary space for the liquid tanks to g14 ¼ VWBT  VWBT
max
0 (30)
maximize the space for the armament. The auxiliary space for the liquid
tanks can be obtained by summing up all volumes of the liquid tanks,
g15 ¼ VLOT
min
 VLOT  0 (31)
including the FOTs, FWTs, WBTs, and LOTs. The third objective function
minimizes the bulkhead damage vulnerability due to an attack in Section
g16 ¼ VLOT  VLOT
max
0 (32)
3.3.1. The bulkhead damage vulnerability (VBD) can be calculated by
using Eq. (3). The objective functions for the optimization of the bulk-
head arrangement are then summarized in Table 6. c Constraints related to the required length for the engine rooms

(4) Constraints If the length of each of the engine rooms should have a minimum
length, such requirements can be formulated as constraints, as shown in
When arranging the bulkheads and decks in the hull, some issues, Eqs. (33)-(34).
including the stability, should be considered. Thus, the following con-
straints were formulated to consider such aspects. g17 ¼ lmin
ER  jx5  x6 j  0 (33)

a Constraints related to the criteria for intact and damage stability g18 ¼ lmin
ER  jx7  x8 j  0 (34)

As mentioned in Section 3.1, a naval surface ship should satisfy the In Eqs. (33)-(34), the values jx5 - x6j and jx7 - x8j are the length of each
criteria for intact and damage stability. The criteria were used as con- of the engine rooms, as shown in Fig. 9.
straints for the optimization in this stage, as shown in Eqs. (17)-(24). Eqs.
(17)-(19) are the criteria for the intact stability, and Eqs. (20)-(24) are the d Constraints related to the required deck height
criteria for the damage stability.
If the height between the adjacent decks should have the minimum
g1 ¼ GZ0 =GZmax  0:6  0 in the intact state (17) distance, such requirements can be formulated as constraints, as shown
in Eqs. (35)-(39).
g2 ¼ 1:4  A2 =A1  0 in the intact state (18)
g19 ¼ hmin
deck  jz1  0j  0 (35)
g3 ¼ ϕ0  15  0 in the intact state (19)
g20 ¼ hmin
deck  jz2  z1 j  0 (36)
g4 ¼ ϕ0  15  0 in the damaged state (20)
g21 ¼ hmin
deck  jz3  z2 j  0 (37)
g5 ¼ 1:4  A2 =A1  0 in the damaged state (21)
g22 ¼ hmin
deck  jz4  z3 j  0 (38)
g6 ¼ 0:075  ðGZmax  HAÞ  0 in the damaged state (22)
g23 ¼ hmin
deck  jD  z4 j  0 (39)
g7 ¼ 0:02  A1  0 in the damaged state (23)
In Eq. (39), the symbol D is the depth of the ship.
Thus, the optimization problem for the bulkhead arrangement has
g8 ¼ Dist  0 in the damaged state (24)

b Constraints related to the required volumes for liquid tanks Table 7

Input data for the room arrangement.

In general, the auxiliary space for the liquid tanks may have required Input data Notation
a given volume for operation. Thus, there are constraints that bind the Hull form and compartment model –
upper and lower limits for each liquid tank, as shown in Eqs. (25)-(32). Number of bulkheads and decks NB&D
Positions of bulkheads and decks from the first stage xi, yj, zk
g9 ¼ VFOT
min
 VFOT  0 (25) Number of rooms NR
Number of passages in the deck i NiP
Locations of entrance doors and inclined ladders See Fig. 11
g10 ¼ VFOT  VFOT
max
0 (26) Adjacency coefficients between the room i and j qij
Longitudinal damage distribution of all threats –
Number of all equipment in the room i NiR
g11 ¼ VFWT
min
 VFWT  0 (27)
Importance factor (between 0 and 1) of the equipment k in the room i Iik
e
Number of the equipment k in the room i Nike
g12 ¼ VFWT  VFWT
max
0 (28) Minimum area for each room amin
i
Number of specific rooms for the fixed location NRfl
Required locations of the specific room i Rreq
i

324
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Fig. 10. Design variables for the optimization of the room arrangement.

Fig. 11. Locations of entrance doors and inclined

ladders.

3.4.2. Formulation for the room arrangement

Table 8
Design variables for the optimization of the room arrangement. Next, the room arrangement in the superstructure is carried out in the
second stage, and several things need to be considered as in the first
Design Description
variables
stage. Regarding the operability and survivability of a naval surface ship,
the sequence of rooms and the positions of the passages in the super-
Ri Room to be allocated at the region i (i ¼ 1, …, NR)
structure are important during the room arrangement. To ensure the
XPik x-coordinate of the passage k in the deck i (from AP, positive
forward) operability, the adjacency index that represents the effectiveness of the
YPik y-coordinate of the passage k in the deck i (from center line, positive flow of crews and supplies in the ship should be minimized. Also, to
towards port side) ensure a sufficient survivability of the ship, the room damage vulnera-
bility due to the attack should be minimized. For these reasons, the
process for the room arrangement was formulated as an optimization
NB&D (¼ 18 in Fig. 9) design variables, 3 objective functions, and 23
problem in this study.
inequality constraints. This problem is a kind of multi-objective, con-
strained optimization problem.
(1) Input data

325
S.-K. Jung et al.
Table 9
Objective functions for the optimization of the room arrangement.
Objective functions Type Mathematical representation
PNR 1 PNR
Adjacency index Minimize F1 ¼ AI ¼
i¼1
PNR
Room damage vulnerability due to Minimize F2 ¼ VRD ¼ i¼1 pi ⋅ei
the attack
When performing a room arrangement, a hull form and a compart-
ment model of the naval surface ship are required, as in the bulkhead
arrangement. Such information includes a number of bulkheads and
decks. In addition, the number of rooms, the number of passages of each
deck, and the locations of the entrance doors and inclined ladders in the
superstructure should also be given, and the positions of the bulkheads
and the decks in the hull that were obtained from the first stage are used
as one of input data in this stage. The locations of entrance doors of rooms
and the locations of inclined ladders play a very important role in
determining the distance (dij) in Eq. (1). Therefore, this study reflected
them, and when calculating the distance between each room, we defined
the distance (dij) so that it can only move through entrance doors and
inclined ladders. The locations of entrance doors and inclined ladders
used as input information in this study are shown in Fig. 11. The locations
of entrance doors and inclined ladders are not included in the design
variables, but since the locations are dependent on the bulkheads and the
passages, they are changed following the related bulkheads and passages
during the optimization process. In the future, we will include them as
design variables of the optimization problem. To evaluate the operability
of the ship, the adjacency coefficients between rooms is needed. To
evaluate the room damage vulnerability due to the attack in Section
3.3.2, the longitudinal damage distribution of all threats, the number of
all equipment in the rooms, and the importance factor and the number of
each equipment in the rooms should also be given. The minimum area for
each room should be input to prevent its value from becoming too small
during the optimization. Finally, the number and the required locations
of the specific rooms (e.g., engine room, bridge, etc.) should be defined.
These input data for the room arrangement are summarized in Table 7.
(2) Design variables
Fig. 12. Components and screenshot of the prototype program for the arrangement design of the naval surface ship.
Ocean Engineering 152 (2018) 316–333
In this stage, the design variables are the sequence of rooms and
positions of passages in the superstructure. Fig. 10 and Table 8 show the
design variables for the optimization of the room arrangement.
In this figure, the number in the square bracket ([ ]) means the region
j¼iþ1 ðqij ⋅dij Þ
ID where each room is to be allocated. Thus, R1 ¼ {OR1}, R2 ¼ {SO1},
R3 ¼ {OR2}, R4 ¼ {SO2}, R5 ¼ {SO3}, …, R33 ¼ {CR10}. That is, the
location of each room can be represented as an array of the room ID (e.g.,
OR1, CR1, etc.) through an encoding process in a genetic algorithm (GA),
and after optimization, the array can be converted into the room
arrangement through the decoding process.
(3) Objective functions
As previously mentioned, it is very important to ensure the opera-
bility and sufficient survivability of the ship in the room arrangement in
the superstructure. Therefore, we introduced two objective functions to
ensure these are obtained in the second stage of the optimization.
The first objective function of this stage minimizes the adjacency
index (AI) in Eq. (1) between the rooms. For this, the distance (dij) be-
tween the rooms should be calculated by using the method in Section
3.2.2. The second objective function minimizes the room damage
vulnerability due to an attack in Section 3.3.2. The room damage
vulnerability (VRD) can be calculated by using Eq. (6). The objective
functions for the optimization of the room arrangement are summarized
in Table 9.
(4) Constraints
When arranging the rooms and passages in the superstructure, some
issues should be considered, including the ship owner's requirements.
Thus, the following constraints were formulated to consider such aspects.
a Constraints related to the required area for rooms
If the area of each room should have the minimum area according to
the ship owner's requirements, such requirements can be formulated as
constraints, as shown in Eq. (40).
326
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 10
Input data for the optimization of the example.

Bulkhead arrangement

Input data Notation Value

Hull form and compartment model – See Fig. 10

Number of bulkheads and decks NB&D 18 (¼ 14 þ 4)
Locations of compartments in the hull, including liquid tanks – See Fig. 9
Required volumes for liquid tanks (FOTs, FWTs, WBTs, and LOTs) Vmin max min max min max min
FOT, VFOT , VFWT, VFWT, VWBT, VWBT, VLOT, 2500 m3, 3000 m3, 80 m3, 110 m3, 800 m3, 1100 m3, 85 m3,
Vmax
LOT 110 m3
Longitudinal damage distribution of all threats – See Fig. 6
Explosive energy per unit mass for all threats EW 4.184⋅109 m2/s2
Charge mass of ASMs, torpedoes, and mines WA, WT, WM 161 kg, 300 kg, 113.4 kg
Minimum length for each of engine rooms lmin
ER 11.5 m
Minimum deck height hmin
deck 1.95 m

Room arrangement

Input data Notation Value

Positions of bulkheads and decks from the first stage x1~x13, y1, z1~z4 Determined from the first stage
Number of rooms NR 33
Number of passages in the deck i NiP 2, 3, 7
Adjacency coefficients between the room i and j qij See Table 11
Locations of entrance doors and inclined ladders - See Fig. 11
Longitudinal damage distribution of all threats – See Fig. 6
Number of all equipment in the room i NiR Different from each room
Importance factor (between 0 and 1) of the equipment k in the Iik
e
room i
Number of the equipment k in the room i Nik
e
Minimum area for each room amin
i
Number of specific rooms for the fixed location NRfl 4
Required locations of the specific room i Rreq
i Rreq req req req
7 ¼ {BR}, R15 ¼ {HC}, R23 ¼ {ER1}, R26 ¼ {ER2}

gi ¼ amin
i  ai  0 ði ¼ 1; …; NR Þ
b Constraints related to the required locations of the specific rooms
If specific rooms, such as the engine room (ER1, ER2), bridge (BR),
and heli cargo (HC), should have fixed locations, such requirements can
be formulated as constraints, as shown in Eq. (41).
 
gNRþi ¼ Ri  Rreq
i ¼ 0 i ¼ 1; …; NRfl
Thus, the optimization problem for the room arrangement has NR (¼
33 þ 12 ¼ 45 in Fig. 10) design variables, 2 objective functions, and
(NR þ NRfl) (¼ 33 þ 4 ¼ 37) constraints. This problem is also a kind of
multi-objective, constrained optimization problem.
3.4.3. Optimization algorithm
Optimization algorithms are divided into global and local optimiza-
tion algorithms, and several types of algorithms are available for global
optimization, including GA (Goldberg, 1989; Davis, 1991). Global opti-
mization algorithms may be suitable for large-scale problems that have
(40) many local optima, but they require more iterations to obtain an
acceptable optimum in comparison with local optimization algorithms.
Several classes of local optimization algorithms also exist, including
sequential linear programming (SLP) (Arora, 2012), sequential quadratic
programming (SQP) (Arora, 2012), and a method for feasible directions
(MFD) (Vanderplaats, 1984). Each of these local optimization algorithms
can be used to effectively find the optimum. However, in some cases,
these local optimization algorithms find an optimum that is relatively
close to the starting point. In addition, the objective functions and con-
(41)
straints presented in this study include nonlinear equations which are
difficult to be solved with general optimization algorithms. Also, since
the permutation is used in the second stage as design variables, the al-
gorithm which can handle the permutation should be used. In this study,
NSGA-II (Deb et al., 2002) which is a kind of multiobjective GA was used
to solve multi-objective nonlinear optimization problems of this study. Of
course, SPEA-II (Zitzler et al., 2001) and SMPSO (Nebro et al., 2009) can
be used. Through comparative tests among them by us, NSGA-II that had
shown the best performance was finally selected and used to solve the
optimization problems of this study.

Table 11
Adjacency coefficients between rooms of the example.

Category – OR CR SB CC EQ ER GA SO COR XO HC COM BR

Officer Room OR 2 – – – – – – – – – – – –
Crew Room CR 1 2 – – – – – – – – – – –
Sick Bay SB 0 2 0 – – – – – – – – – –
Command & Control Room CC 2 0 0 0 – – – – – – – – –
Equipment Storage EQ 1 2 0 2 2 – – – – – – – –
Engine Room ER 1 0 1 0 2 2 – – – – – – –
Galley GA 0 3 2 0 1 2 0 – – – – – –
Ship Office SO 0 2 0 0 0 0 0 2 – – – – –
Commander Room COR 2 0 0 3 0 2 0 0 0 – – – –
Executive Office Room XO 2 0 0 3 0 2 0 0 3 0 – – –
Heli Cargo HC 1 2 0 0 0 0 0 0 0 0 0 – –
Communication Room COM 2 0 0 0 0 0 0 0 0 0 0 0 –
Bridge BR 2 0 0 0 1 0 0 0 0 0 0 1 0

327
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 12
Comparison of the objective function values for the bulkhead arrangement between the initial and first optimal designs (large influence of damage vulnerability).

Objective functions Type Initial (A) Optimization (B) Ratio (B/A-1)

Maximum bending moment at the intact state Minimize 362,261 kN⋅m 321,860 kN⋅m 11.1%
Auxiliary space for liquid tanks Minimize 4390 m3 3823 m3 12.9%
Bulkhead damage vulnerability due to the attack Minimize 13.81 GPa 6.89 GPa 50.1%

Table 13
Comparison of the values of design variables for the bulkhead arrangement between the initial and first optimal designs.

Design variables Initial (A) Optimization (B) Difference (B-A) Design variables Initial (A) Optimization (B) Difference (B-A)

x1 10.2 11.6 1.3 x10 116.8 117.3 0.5

x2 23.0 22.8 0.2 x11 128.6 128.6 0.0
x3 33.6 34.8 1.2 x12 140.9 141.9 1.0
x4 46.5 47.1 0.6 x13 52.4 53.1 0.7
x5 57.8 58.2 0.4 y1 4.7 4.8 0.1
x6 70.2 69.7 0.5 z1 3.1 2.8 0.3
x7 80.8 80.9 0.1 z2 5.3 5.0 0.3
x8 93.9 92.5 1.4 z3 9.3 9.3 0.0
x9 104.4 105.0 0.6 z4 12.2 12.2 0.0

4. Application of the proposed method for the naval surface ship
4.1. Prototype program
A prototype program was developed using the proposed method in
the C# programing language, and it consists of the template model,
stability module, operability module, survivability module, optimization
module, and user interface, as shown in Fig. 1. Fig. 12 shows the com-
ponents and the screenshot of the prototype program. To verify the
applicability of the proposed method and the prototype program, they
were applied to an example of the arrangement design of the naval sur-
face ship.
4.2. Description of an example
As mentioned earlier, the proposed method focuses on the arrange-
ment design of the naval surface ship. The bulkhead arrangement was
optimized in the first stage of optimization, and in the second stage, the
room arrangement was optimized. A 7000-ton class missile destroyer was
selected as an example in this study. The length, breadth, and draft of the
ship are 154 m, 20 m, and 9.4 m, respectively. The bulkhead arrangement
in the hull, including tanks, is shown in Fig. 9 and the room arrangement
in the superstructure is shown in Fig. 10.
In this study, we optimized the arrangement of 14 bulkheads and 4
decks during the first stage. After optimizing the bulkhead arrangement,
we have optimized the arrangement of 33 rooms and 12 passages, and we
then present each result of the arrangement. More details of the input
data are summarized in Table 10. Table 11 shows the adjacency co-
efficients (qij) between the rooms of the example.
4.3. Optimization of the bulkhead arrangement
The arrangement of the 14 bulkheads and 4 decks was optimized in
the first stage, as mentioned earlier. In this stage, the positions of the
bulkheads and the decks were optimized to minimize the maximum
bending moment at the intact state, to minimize the auxiliary space for
liquid tanks, and to minimize the bulkhead damage vulnerability due to
the attack, as mentioned in Section 3.4.1.
Three objective functions were used to optimize the bulkhead
arrangement. Therefore, three-dimensional Pareto-optimal set can be

Fig. 13. Comparison of the bulkhead arrangement between the initial and first optimal designs.

328
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 14
Comparison of the objective function values for the bulkhead arrangement between the Table 16
initial and second optimal designs (even influence of three objective functions). Comparison of the objective function values for the bulkhead arrangement between the
second optimal designs without and with post-processing.
Objective functions Type Initial (A) Optimization Ratio (B/
(B) A-1) Objective functions Type w/o Post- w/Post- Ratio
processing (A) processing (B) (B/A-1)
Maximum bending Minimize 362,261 kN⋅m 315,058 kN⋅m 13.0%
moment at the Maximum bending Minimize 315,058 kN⋅m 330,857 kN⋅m þ5.0%
intact state moment at the
3 3
Auxiliary space for Minimize 4390 m 3813 m 13.1% intact state
liquid tanks Auxiliary space for Minimize 3813 m3 3824 m3 þ0.3%
Bulkhead damage Minimize 13.81 GPa 11.61 GPa 15.9% liquid tanks
vulnerability due Bulkhead damage Minimize 11.61 GPa 12.85 GPa þ10.7%
to the attack vulnerability due
to the attack

obtained as a result of optimization. In this section, we compared the

initial design and two optimal designs. As the first optimal design, the
design alternative with a large influence of damage vulnerability which is
an important factor in the bulkhead arrangement was selected. As the
second optimal design, the design alternative evenly considering all the
effects of the objective functions was selected through the normalization
of three objective functions. Table 12 shows a comparison of the objec-
tive function values for the bulkhead arrangement between the initial
design and the first optimal design (large influence of damage vulnera-
bility). In the case of the initial design, the values of the objective func-
tions were calculated to be the same as the optimal design. That is, they
were calculated according to the formulation mentioned in Section 3.4.1.
As shown in Table 12, the maximum bending moment in the intact
state decreased from 362,261 kN⋅m (initial design) to 321,860 kN⋅m
(first optimal design) by 11.1%. Also, the auxiliary space for the liquid
tanks decreased from 4390 m3 (initial design) to 3823 m3 (first optimal
design) by 12.9%. The bulkhead damage vulnerability improved signif-
icantly from 13.81 GPa (initial design) to 6.89 GPa (first optimal design)
by 50.1%. Further comparisons between the initial design and the first
optimal design for the bulkhead arrangement are shown in Table 13 and
Fig. 13.
The bulkhead damage vulnerability depends on the summation of the
explosion pressures (Pir) of the bulkheads and decks due to explosives
such as ASMs, torpedoes, and mines, as shown in Eq. (3). From Eqs. (4)
and (5), we can see that the explosion pressure of the bulkhead or deck is
inversely proportional to its distance from the explosion point. Mean-
while, the distribution of the explosion points by the ASMs and torpedoes
has its peak point near the midship, as shown in Fig. 6. The distribution
by the mines has its peak point near fore body. Therefore, to minimize the

Table 15
Comparison of the values of design variables for the bulkhead arrangement between the initial and second optimal designs.

Design variables Initial (A) Optimization (B) Difference (B-A) Design variables Initial (A) Optimization (B) Difference (B-A)

x1 10.2 10.2 0.0 x10 116.8 117.2 0.4

x2 23.0 23.1 0.1 x11 128.6 128.5 0.1
x3 33.6 35.0 1.4 x12 140.9 141.9 1.0
x4 46.5 45.3 1.2 x13 52.4 52.2 0.2
x5 57.8 58.3 0.6 y1 4.7 4.9 0.2
x6 70.2 70.6 0.4 z1 3.1 2.7 0.4
x7 80.8 80.9 0.1 z2 5.3 5.0 0.3
x8 93.9 93.1 0.8 z3 9.3 9.3 0.0
x9 104.4 104.4 0.0 z4 12.2 12.2 0.0

Fig. 14. Comparison of the bulkhead arrangement between the initial and the second optimal designs.

329
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 17
Comparison of the values of design variables for the bulkhead arrangement between the second optimal designs without and with post-processing.

Design variables Optimization (A) Post processing (B) Difference (B-A) Design variables Optimization (A) Post processing (B) Difference (B-A)

x1 10.2 10.0 0.2 x10 117.2 117.0 0.2

x2 23.1 23.0 0.1 x11 128.5 128.5 0.0
x3 35.0 35.0 0.0 x12 141.9 142.0 0.1
x4 45.3 45.5 0.2 x13 52.2 52.0 0.2
x5 58.3 58.5 0.2 y1 4.9 5.0 0.1
x6 70.6 70.5 0.1 z1 2.7 2.5 0.2
x7 80.9 81.0 0.1 z2 5.0 5.0 0.0
x8 93.1 93.0 0.1 z3 9.3 9.5 0.2
x9 104.4 104.5 0.1 z4 12.2 12.0 0.2

Table 18 13.1%. The bulkhead damage vulnerability improved from 13.81 GPa
Comparison of the objective function values for the room arrangement between the initial (initial design) to 11.61 GPa (second optimal design) by 15.9%. Further
and first optimal designs (large influence of adjacency index).
comparisons between the initial design and the second optimal design for
Objective functions Type Initial Optimization Ratio (A/ the bulkhead arrangement are shown in Table 15 and Fig. 14.
(A) (B) B-1)
In the second optimal design, the distances between the bulkheads
Adjacency index Minimize 21,976 16,359 25.6% near the midship were narrowed as compared with the initial design, as
Room damage Minimize 0.134 0.127 5.2% shown in Fig. 14. They caused the reduction of the capacity of liquid
vulnerability due to the
attack
tanks. As a result, the weight of the midship was reduced and the bending
moment was reduced accordingly. We can see that the maximum bending
moment at the intact state and the auxiliary space for liquid tanks of the
bulkhead damage vulnerability, the distances between the bulkheads second optimal design are slightly better than those of the first optimal
near the midship and fore body were changed during the optimization. design, as shown in Tables 12 and 14. That is, the improvement of them
That is, in the case of the midship, the distances between x6 and x7 and were not so much as compared with the improvement of the bulkhead
between x8 and x9 increased. As a result, the distances between x5 and x6 damage vulnerability of the first optimal design. This is because the
(1st engine room) and between x7 and x8 (2nd engine room) decreased. bulkhead damage vulnerability is very sensitive to the positions of the
In the case of the fore body, the distance between x11 and x12 increased. bulkheads.
However, the distance between x10 and x11 decreased. In practice, the bulkheads should be located at the frames of the ship
Table 14 shows a comparison of the objective function values for the to take into account manufacturability. To reflect this, the post-
bulkhead arrangement between the initial design and the second optimal processing process was done to move the bulkheads of the optimal de-
design (even influence of three objective functions). As shown in signs to the adjacent frames. The frame spacing of the naval surface ship
Table 14, the maximum bending moment in the intact state decreased used in this example is assumed to be 0.5 m. Thus, this process means
from 362,261 kN⋅m (initial design) to 315,058 kN⋅m (second optimal moving the optimal positions of the bulkheads to the nearest frame of
design) by 13.0%. Also, the auxiliary space for the liquid tanks decreased 0.5 m spacing. This process was applied to the second optimal design and
from 4390 m3 (initial design) to 3813 m3 (second optimal design) by the results without and with this process were compared.

Table 19
Comparison of the values of design variables for the room arrangement between the initial and first optimal designs.

Design variables Initial (A) Optimization (B) Difference (B-A) Design variables Initial (A) Optimization (B) Difference (B-A)

YP11 0.0 0.3 0.3 YP32 5.0 5.6 0.6

YP12 0.0 0.4 0.4 YP33 0.0 0.5 0.5
YP21 0.0 0.1 0.1 YP34 5.0 4.2 0.8
YP22 3.0 2.4 0.6 YP35 5.0 5.6 0.6
YP23 3.0 3.8 0.8 XP31 41.0 42.2 1.2
YP31 5.0 4.2 0.8 XP32 110.0 110.3 0.3

Fig. 15. Comparison of the room arrangement between the initial and first optimal designs.

330
S.-K. Jung et al. Ocean Engineering 152 (2018) 316–333

Table 20 attack, as mentioned in Section 3.4.2.

Comparison of the objective function values for the room arrangement between the initial
and second optimal designs (even influence of two objective functions).
Objective functions Type Initial Optimization
(A) (B)
Adjacency index Minimize 21,976 18,435
Room damage Minimize 0.134 0.114
vulnerability due to the
attack
Table 16 shows a comparison of the objective function values for the
bulkhead arrangement between the second optimal designs without and
with post-processing. As shown in Table 16, the maximum bending
moment in the intact state increased from 315,508 kN⋅m (without post-
processing) to 330,857 kN⋅m (with post-processing) by 5.0%. Also, the
auxiliary space for the liquid tanks increased from 3813 m3 (without
post-processing) to 3824 m3 (with post-processing) by 0.3%. The bulk-
head damage vulnerability deteriorated from 11.61 GPa (without post-
processing) to 12.85 GPa (with post-processing) by 10.7%. Further
comparisons between the second optimal designs without and with post-
processing for the bulkhead arrangement are shown in Table 17.
As a result of comparing the objective functions of the two design, the
maximum bending moment at the intact state and the auxiliary space for
liquid tanks showed a similar or slight increase. However, as described
above, the bulkhead damage vulnerability due to the attack was very
sensitive to the positions of the bulkheads so that the amount of increase
was greater than the others. If we take some disadvantage in the objective
functions, we can enhance manufacturability at the same time. There-
fore, the post-processing process is necessary for practical design.
4.4. Optimization of the room arrangement
Next, the arrangement of 33 rooms and 12 passages was optimized in
the second stage. In this stage, the sequence of rooms and the positions of
the passages were optimized to minimize the adjacency index between
the rooms and to minimize the room damage vulnerability due to the
Two objective functions were used to optimize the room arrange-
ment. Therefore, two-dimensional Pareto-optimal set can be obtained as
Ratio (B/ a result of optimization. In this section, we compared the initial design
A-1)
and two optimal designs. As the first optimal design, the design alter-
16.1% native with a large influence of adjacency index which is an important
14.9% factor in the room arrangement was selected. As the second optimal
design, the design alternative evenly considering all the effects of the
objective functions was selected through the normalization of two
objective functions. Table 18 shows a comparison of the objective func-
tion values for the room arrangement between the initial design and the
first optimal design (large influence of adjacency index). When per-
forming the optimization, the adjacency information for the room
arrangement was extracted through a consultation with experts of naval
surface ships. In the case of the initial design, the values of objective
functions were calculated to be the same as the optimal design. That is,
they were calculated according to the formulation mentioned in Section
3.4.2.
As shown in Table 18, the adjacency index of the optimal design
decreased from 21,976 (initial design) to 16,359 (first optimal design) by
25.6%, and the room damage vulnerability improved slightly from 0.134
(initial design) to 0.127 (first optimal design) by 5.2%. More compari-
sons between the initial design and the first optimal design for the room
arrangement are shown in Table 19 and Fig. 15.
According to the adjacency coefficients between rooms in Table 11,
the galley (GA) and the crew rooms (CR1~CR10), the commander room
(COR) and the command & control room (CC), the executive office room
(XO) and the CC, and the XO and the COR have a high value (¼ 3) of the
adjacency coefficient between them. The GA and the ER, the COR and the
ER, and the XO and the ER have a low value (¼ 2) of the adjacency
coefficient between them. To minimize the adjacency index, the rooms
with a high value of the adjacency coefficient should be arranged close to
each other and the rooms with a low value should be arranged apart from
each other. In Fig. 15, we can see that the optimal design satisfied these
requirements well, relative to the initial design.
The room damage vulnerability depends on the summation of the

Table 21
Comparison of the values of design variables for the room arrangement between initial and second optimal designs.

Design variables Initial (A) Optimization (B) Difference (B-A) Design variables Initial (A) Optimization (B) Difference (B-A)

YP11 0.0 0.6 0.6 YP32 5.0 4.2 0.8

YP12 0.0 0.4 0.4 YP33 0.0 0.7 0.7
YP21 0.0 0.1 0.1 YP34 5.0 5.8 0.8
YP22 3.0 3.0 0.0 YP35 5.0 4.2 0.8
YP23 3.0 2.6 0.4 XP31 41.0 42.4 1.4
YP31 5.0 5.8 0.8 XP32 110.0 109.8 0.2

Fig. 16. Comparison of the room arrangement between the initial and second optimal designs.

331
S.-K. Jung et al.
Table 22
Comparison of the objective function values for the room arrangement between the second
optimal designs without and with adjustment.
Objective functions Type w/o w/ Ratio
adjustment adjustment Ratio (B/
(A) (B) A-1)
Adjacency index Minimize 18,435 19,595 þ6.3%
Room damage Minimize 0.113 0.124 þ9.7%
vulnerability due to
the attack
multiplication of the probability of damage (pi) due to explosives and the
importance factor (ei) of rooms, as shown in Eq. (6). The probability of
damage of the rooms is proportional to the multiplication of the proba-
bility of damage length (pDL) and the probability of impact (pIP), as shown
in Eq. (7). To minimize the room damage vulnerability, the rooms (e.g.,
command & control room, equipment storages) with a high importance
factor should be arranged apart from the explosive points as far as
possible. Since most of the explosive points are located at midship and at
the fore body, such rooms should be arranged away from the points. As
shown in Fig. 15, in the case of the initial design, some of the equipment
storage (EQ2 and EQ3) is located midship. However, we can see that such
rooms were arranged aft and forward in the optimal design. In addition,
the CC was also located in the fore body but after optimization we can see
that it moved to after body. As a result, the room damage vulnerability of
the optimal design was lower than that of the initial design.
Table 20 shows a comparison of the objective function values for the
room arrangement between the initial design and the second optimal
design (even influence of two objective functions). As shown in Table 20,
the adjacency index decreased from 21,976 (initial design) to 18,435
(second optimal design) by 16.1%, and the room damage vulnerability
improved slightly from 0.134 (initial design) to 0.114 (second optimal
design) by 14.9%. Unlike the previous results, it can be seen that the two
objective functions were improved at a similar rate. More comparisons
between the initial design and the optimal design for the room
arrangement are shown in Table 21 and Fig. 16.
In the second optimal design, the equipment storage (EQ4) moved to
the after body, as shown in Fig. 16. We can see that the vulnerability of
the second optimal design is slightly better than that of the first optimal
design. This is because explosive points are concentrated relatively in the
midship and the fore body.
If we look closely at the optimization results, some parts of the
optimal design would be quickly challenged by an experienced designer.
For example, it is hard to believe that anyone would place crew rooms on
the O-2 deck just after of the bridge. This is a serious work area of the
ship, not a rest and recreation area. The crew wants to be away from
Fig. 17. Comparison of the room arrangement between the second optimal designs without and with adjustment.
332
Ocean Engineering 152 (2018) 316–333
officers and close to the mess decks when off duty. Further, it appears that
the area of the command & control room (CC) has been shrunk and
personnel access distance from the bridge to the CC has been more than
doubled by the optimization process.
To reflect these considerations, we adjusted several adjacency co-
efficients for calculating the adjacency index. The adjacency coefficient
between the bridge (BR) and the commander room (COR) was adjusted
to 3 to make it easier for the captain to access the bridge. Similar reason,
the adjacency coefficient between the BR and the executive officer room
(XO) was adjusted to a high value (¼ 3). And other adjacency coefficients
was adjusted so that crews and officers could be away from the opera-
tional area during off duty. That is, the adjacency coefficient between the
CC and the crew room (CR), the CC and officer room (OR), the BR and the
CR, and the BR and the OR were adjusted to a low value (¼ 2). In
addition, constraints were added so that the CR would not be located at
O-2 deck as similar with Eqs. (35)-(39).
Table 22 shows a comparison of the objective function values for the
room arrangement between the second optimal designs without and with
adjustment of adjacency coefficients and constraints. As shown in
Table 22, the adjacency index increased from 18,435 (without adjust-
ment) to 19,595 (with adjustment) by 6.3%, and the room damage
vulnerability slightly deteriorated from 0.113 (without adjustment) to
0.124 (with adjustment) by 9.7%. By changing the adjacency coefficients
and constraints, we can see that both objective function values were
worse than the second optimal design. More comparisons between the
second optimal designs without and with adjustment for the room
arrangement are shown in Fig. 17.
Due to the change of adjacency coefficients and constraints, the dis-
tances between the COR and the CC, the BR and the CC, and the BR and
the XO were closer. The COR and the XO moved to the upper deck of the
fore body. In addition, all of the CR were placed on the main deck so that
the optimal design similar to an actual naval surface ship was obtained.
5. Conclusion and future studies
This study proposed an arrangement method for a naval surface ship.
In the preliminary design stage, it is not easy to consider all requirements
for the performance parameters of the ship, including the stability,
operability, and survivability, in a short period of time. An optimization
technique was introduced here to effectively solve such problem. First, a
template model is defined as the data structure to represent the
arrangement information of the ship. Second, methods were studied to
quantitatively evaluate the stability, operability, and survivability of the
ship, by referring existing studies. Third, two-stage optimization prob-
lems for the arrangement design were mathematically formulated with
the selection of the design variables, objective functions, and constraints.
S.-K. Jung et al.
Finally, a user interface was developed to input the given data and to
output the optimization results for the arrangement design. To check the
applicability of the proposed method, a prototype program based on the
proposed method was developed, and then the method was applied to an
example of a 7000-ton class missile destroyer. The results show that the
method could yield a better alternative for the arrangement design in
terms of the objective functions that were selected for the optimization.
In the future, a mathematical formulation of the optimization module
will be further refined to obtain a more realistic arrangement design.
Also, the survivability module will be improved to consider various as-
pects of the survivability besides the vulnerability. Further consider-
ations of the interface between the hull and the superstructure of the ship
will be made for the optimization process in the future. To increase the
applicability of the proposed method, it will be applied to various types
of naval surface ships.
Acknowledgments
This study was partially supported by (a) The National Research
Foundation of Korea (NRF) grant funded by the Ministry of Education of
the Republic of Korea (No. 2016R1A2B4016253), (b) BK21 Plus Program
(Education and Research Center for Creative Offshore Plant Engineers)
funded by the Ministry of Education of the Republic of Korea, and (c)
Research Institute of Marine Systems Engineering of Seoul National
University, Republic of Korea.
References
Andrews, D.J., 2006. Simulation and the design building block approach in the design of
ships and other complex systems. Proc. R. Soc. Lond. A: Math. Phys. Eng. Sci. 462
(2075), 3407–3433.
Arora, J.S., 2012. Introduction to Optimum Design, third ed. Academic Press, Iowa, USA.
Boulougouris, E.K., Papanikolaou, A.D., 2004. Optimization of the survivability of naval
ships by genetic algorithms. In: Proceedings of the International Conference on
Computer and IT Applications and in the Maritime Industries 2004, Siguenza, Spain,
pp. 175–189, 2004.05.09-12.
Brown, A.J., Deybach, F., 1998. Towards a rational intact stability criteria for naval ships.
Nav. Eng. J. 110, 65–77.
Byun, Y.C., 1998. A Study on Determination of Ship Main Dimension and Compartments
Arrangement Supporting Expert System. M.Sc. thesis. Seoul National University,
Republic of Korea.
Choi, Y.S., Yeo, S.K., Chung, H., 2015. An optimal arrangement method of blast hardened
bulkhead for surface warships using a simplified vulnerability evaluation. In:
Proceedings of the 2015 World Congress on Advances in Structural Engineering and
Mechanics, Incheon, Republic of Korea, pp. 1–19, 2015.08.25-29.
Chung, B.Y., Kim, S.Y., Shin, S.C., Koo, Y.H., Kraus, A., 2011. Optimization of
compartments arrangement of submarine pressure hull with knowledge based
system. International Journal of Naval Architecture and Ocean Engineering 3 (4),
254–262.
Davis, L., 1991. Handbook of Genetic Algorithms. Van Nostrand-Reinhold, New York,
USA.
Deb, K., Pratap, A., Agarwal, S., Meyarivan, T., 2002. A fast and elitist multiobjective
genetic algorithm: NSGA-II. IEEE Trans. Evol. Comput. 6 (2), 182–197.
Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization, and Machine Learning.
Addison-Wesley, Reading, USA.
333
Ocean Engineering 152 (2018) 316–333
Harmsen, E., Krikke, M., 2000. A probabilistic damage stability calculation method for
naval vessels. In: Proceedings of the International Conference on Stability of Ships
and Ocean Vehicles 2000, Australia, Tasmania, pp. 330–350, 2000.02.07-12.
Helvacioglu, S., Insel, M., 2005. A reasoning method for a ship design expert system.
Expet Syst. 22 (2), 72–77.
Ju, S.H., Kim, Y.M., Jeong, Y.K., Shin, J.G., 2014. Compartment arrangement algorithm
and application based on the survivability of naval vessel. In: Proceedings of the
Society of CAD/CAM Conference 2014, Pyeongchang, Republic of Korea,
pp. 1167–1170, 2014.02.12-14.
Jung, J.W., 2016. Optimization of Spatial Arrangement at an Early Design Stage of Naval
Ship through Layer-based Integrated Survivability Evaluation. M.Sc. thesis. Seoul
National University, Republic of Korea.
Kim, K.S., Roh, M.I., 2016. A submarine arrangement design program based on the expert
system and the multistage optimization. Adv. Eng. Software 98, 97–111.
Kim, K.S., Roh, M.I., Ha, S., 2015. Expert system based on the arrangement evaluation
model for the arrangement design of submarine. Expert Syst. Appl. 42, 8731–8744.
Kim, J.S., Roh, M.I., Ham, S.H., 2017a. A method for intermediate flooding and sinking
simulation of a damaged floater in time domain. J. Comput. Des. Eng. 4 (1), 1–13.
Kim, S.K., Roh, M.I., Kim, K.S., 2017b. Arrangement method of offshore topside based on
an expert system and optimization technique. J. Offshore Mech. Arctic Eng. 139 (2),
1–19.
Lee, K.Y., Han, S.N., Roh, M.I., 2003. An improved genetic algorithm for facility layout
problems having inner structure walls and passages. Comput. Oper. Res. 30 (1),
117–138.
Lee, K.Y., Roh, M.I., Jeong, H.S., 2005. An improved genetic algorithm for multi-floor
facility layout problems having inner structure walls and passages. Comput. Oper.
Res. 32 (4), 879–899.
Moore, C.S., Paulling, J.R., 2010. The Principles of Naval Architecture Series: Intact
Stability. The Society of Naval Architects and Marine Engineers, USA.
Republic of Korea Navy, 2014. Criteria for Stability. Republic of Korea Navy, Republic of
Korea.
Nebro, A., Alba, E., Molina, G., Chicano, F., Luna, F., Durillo, J., 2009. SMPSO: a new
PSO-based metaheuristic for multi-objective optimization. In: Proceedings of the
2009 IEEE Symposium on Computational Intelligence in Multicriteria Decision-
making (MCDM 2009), Ottawa, Canada, pp. 66–73, 2009.07.08-10.
Nick, E.K., 2008. Fuzzy Optimal Allocation and Arrangement of Spaces in Naval Surface
Ship Design. Ph.D. thesis. The University of Michigan, USA.
Oers, B.K., Stapersma, D., Hopman, H., 2008. Issues when selecting naval ship
configurations from a Pareto-optimal set. In: Proceedings of the 12th AIAA/ISSMO
Multidisciplinary Analysis and Optimization Conference, Victoria, Canada,
2008.09.10–12.
Parsons, M., Chung, H., Nick, E., Daniels, A., Liu, S., Patel, J., 2008. Intelligent Ship
Arrangements (ISA): a new approach to general arrangement. Nav. Eng. J. 120 (3),
51–65.
Przemieniecki, J.S., 1994. Mathematical Methods in Defense Analyses, second ed.
American Institute of Aeronautics and Astronautics, USA.
Roh, M.I., Lee, S.U., Lee, K.Y., 2009. Development of an optimal compartment design
system of naval ships using compartment modeling and ship calculation modules.
Journal of the Society of Naval Architects of Korea 46 (4), 424–434.
Said, M.O., 1995. Theory and practice of total ship survivability for ship design. Nav. Eng.
J. 107 (3), 191–203.
Sarchin, T.H., Goldberg, L.L., 1962. Stability and buoyancy criteria for U.S. naval surface
ships. SNAME Transactions 70, 418–458.
Shin, J.H., 2013. A Study on the Spatial Arrangement of Naval Ships Considering
Survivability. M.Sc. thesis. Seoul National University, Republic of Korea.
Shin, S.C., Kim, S.Y., Park, J.K., 2002. Evaluation of engine room machinery arrangement
using fuzzy modeling. Journal of Fuzzy Logic and Intelligent Systems 12 (2),
157–163.
US Department of the Army, 1990. Structures to Resist the Effects of Accidental
Explosions, TM5–1300. US Department of the Army, USA.
Vanderplaats, G., 1984. Numerical Optimization Techniques for Engineering Design.
McGraw-Hill, New York, USA.
Zitzler, E., Laumanns, M., Thiele, L., 2001. SPEA2: improving the strength Pareto
evolutionary algorithm for multiobjective optimization. In: Proceedings of the
EUROGEN 2001 Conference, Athens, Greece, pp. 95–100, 2001.09.19-21.