Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Papandreouch Swath

Download as pdf or txt
Download as pdf or txt
You are on page 1of 108

NATIONAL TECHNICAL UNIVERSITY OF ATHENS

SCHOOL OF NAVAL ARCHITECTURE AND MARINE


ENGINEERING
DIVISION OF SHIP DESIGN AND MARITIME TRANPORT

PARAMETRIC DESIGN AND MULTI-


OBJECTIVE OPTIMIZATION OF SWATH

DIPLOMA THESIS
OF
PAPANDREOU
CHRISTOS

SUPERVISOR PROFESSOR : A. D. PAPANIKOLAOU

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

1
Contents……………………………………………………………………………2
Acknowledgments………………………………………………………………..4
Introduction………………………………………………………………………. 4

1. CAD/CAE systems for ship design…………………………..….……...…. 7


 1.1Introduction……………………..……………………………….….…. 7
 1.2 Objectives of CAD and CAE applied to hull forms………..........….7
 1.3 Ship design process……………………………………………….… 8
 1.4 An insight in fully-parametric ship design……………………..…. 10
 1.5 FRIENDSHIP-FRAMEWORK…………………..………………..... 12
o 1.5.1 Design principles………………………………………………... 12
o 1.5.2 Basic elements…………………………………………………... 13
 1.5.2.1 Metasurfaces………………………………………….......... 13
 1.5.2.2 Feature modeling…………………………………….......... 13
 1.5.2.3 Curve Engine……………………………………………...... 14
 1.5.2.4 Design Engine…………………………………………….... 15

2. Optimization ………………………………………………………………….. 16
 2.1 Introduction-The generic ship design optimization problem…….. 16
 2.2 Single Objective & Multi-Objective optimization…….…..……...… 17
 2.3 Generic optimization problem……………………………………..... 18
 2.4 Solving a multi-objective optimization problem…………...………..19
 2.5 Design of Experiment………………………………………….…….. 20
 2.6 Genetic Algorithms (GA)……………………………………….……. 21

3. Multi hull ships and SWATH………………………………………….25


 3.1 Multi hull ships………………………………………………..………25
 3.2 General aspects of Swaths…………………………………….....…26
 3.3 Swath design parameters……………………………………………27
 3.4 Comparison of Swath with conventional ships…………………...29
 3.5 History of Swath……………………………………………………….29
 3.6 New built Swath………………………………………………..…...…36

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

2
4. The calculation of ship’s resistance of multi-hull vessels……...42
 4.1 Components of ship‟s resistance……….……..…………………42
 4.2 Thin ship theory……………………………………..………………43
 4.3 The code MICHLET…………………………………………………44
 4.4 Basic considerations…...…………………………..………………45
 4.5 Calculation of Resistance………………………...…………...…..46
 4.6 Input files………..……………………………………...……………46
 4.7 General input variables…..………………………………...………47

5. Case Study – Design and optimization of AQ3…………….…….53


 5.1 Design of AQ3………………………………………………….……..53
o 5.1.1 Parametric model design………………………………….53
o 5.1.2 Initial parameters……………………………………………55
o 5.1.3 Final parameters…………………………………………….58
 5.2 Optimization procedure……………………………………....……….58
 5.3 Hydrostatics…………………………………………………………….59
 5.4 MICHLET integration…………………...……………………..………61
 5.5 Evaluations of optimization…………………………………...………63
o 5.5.1 Design of experiment (DOE)……………………………….65
o 5.5.2 Multi objective optimization-Genetic algorithms……….…74
 5.6 Best designs……………………………………………………………81
 5.7 Validating results with SHIPFLOW………………………..…………88
 5.8 Single-objective optimization…………………………………………94
 5.9 Comparison of AQ3 with AQ………………………………………....96

6. Conclusions……………………………………………………….…….97

7. References……………………………………………………………….99

Appendices………………………………………………………..………101
 Appendix A - TriSwath…………………………………………….…….101
 Appendix B - DiamontSwath………………………………..………….105

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

3
ACKNOWLEDGMENT

At this point I would like to express my deepest gratitude to my advisor Prof.


A. D. Papanikolaou for giving me the opportunity to deal with such an
interesting topic and for his decisive support and motivation.

My sincere thanks also go to Dr. E. Boulougouris for his assistance during the
completion of this Diploma thesis, to Mr. D. Mourkogiannis for his great
support with CFD-configurations and of course to my colleagues G.
Koutroukis and T. Plessas for their help to any query turned up.

I am also thankful to Mr. Konrad Lorentz , Mr. Jörg Palluch , Mr. Arne
Bergmann , Mr. Karsten Wenzke , Stefan Wunderlich , Mr. Ben Zeitz , for
giving me valuable help and guidance while working with FRIENDSHIP-
FRAMEWORK.

In addition I like to deeply thank Dr. Leo Lazauskas for providing me the code
Michlet free of charge for the conduct of my thesis. Also, I like to express my
appreciation to the work of late Professor E. O. Tuck on "thin ship theory",
with whom Dr. Lazauskas closely worked for many years.

Last but not least, I would like to thank my family and friends for their great
support in all this period.

INTRODUCTION

Shipping has been one of the earliest activities of the human race, as a result
of the continuous struggle to survive, thrive, expand and explore. Throughout,
the known history kingdoms, empires and economies have heavily depended
on the sea and maritime transportation not only in the mean of conflict but
more important in the means of communication, cooperation in the forms of
trade and commerce. In the meantime, during the evolution of the scientific
thinking and technology, shipping has been affected with several
technological breakthroughs such as the introduction of sail, steam, iron and
steel, welding procedures, diesel engines and many more.

While the ship technology advanced so did the operations and the support
from a vast regulatory FRAMEWORK which initially depended solely on
experience and mutual agreements but then was rationalized with the
introduction of engineering and Naval Architectural principles.

Today, shipping is a very complex and volatile organism which is triggered


and influenced by several exogenous organisms such as supply and demand

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

4
functions, global conflicts and economy as well as ever increasing and
sensitive oil prices. The ship designer, shipwright, owner, operator and
charterer of today are challenged more than ever to survive not only due to
the globalization and the increased level of competence but also due to the
ever increasing order of volatility and uncertainty in global economic markets.

For this particular reason, when looking from a technical point of view, the
Naval Architect of today is responsible not only for designing , delivering and
operating a superior product but for making that product optimum in terms of
efficiency, safety and more recently in terms of environmental friendliness.
This term characterizes the 21st century, as it is a result of an excessive
pressure from the society towards the policy makers to make steps towards a
more sustainable and green profile in all levels of human activities including
shipping.

When responsible for such a complex and difficult procedure the Naval
Architect must be analytical and follow the principles of holism which is
reflected in Holistic Ship design and Operation. Holistic Ship Design uses
principles of Multi-Objective Optimization in order to solve conflicting
arguments towards the best solution which satisfies the user requirements
while it complies with the constraints set by society in form of Rules and
Regulations.
This present thesis aims at giving a solution and a tool to the Naval Architect
of today, and in particular those interested in Swath Design. Swaths are ideal
for optimization since as ship systems usually they include a lot of
contradicting requirements while the consequences of a potential accident
can be catastrophic(referring to passenger ships).

More specific at this thesis chapter 1 is generally about CAD/CAE systems


for ship design. It is described how are used , what they can do and there is a
description about the software exploited in the present thesis.

Chapter 2 is relevant with optimization and how it works. It analyses the


optimization problem , separates the single from the multi objective
optimization and describes the genetic algorithms.

Chapter 3 talks about multi-hull ships at general and especially Swaths. It


quotes the aspects the design parameters the history of Swath. In addition it
describes the comparison of them with conventional ships and present the
newly built Swath.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

5
Chapter 4 contains the case study of the Swath design and optimization(AQ3)
and it is obvious the core of the thesis. It is widely acknowledged that at
present times an economic option is very attractive to the public, especially if
we talk about passenger and car ships. So regarding to this, the optimization
was made in order to minimize every aspect of resistance. As a result some
designs were found for every need. Either the best model for a high speed
that serves the public at summers or the best model at a relatively low speed
that that serves at a non commercial period, or a optimized model for a
combination of the above. All these accompanied with the strict regulations
and safety rules.
All of these developments create a new (almost chaotic) patch that both the
existing and new built ships have to respond to. It is generally admitted that
these requirements change entirely the way we think about shipping in
general and the challenge is big and often a handicap and burden for the
operability and profitability of the owning and managing company. This
creates a new need for innovative, safer and more efficient designs that will
not mitigate the economic performance, sustainability and competitiveness of
each concept. Initially though these designs are not easily acceptable by
shipyards and need additional capital expenditure.
They key in making them a sustainable option is reduce the Operating
Expenditures (OPEX) and maintenance costs, as well increase the availability
and reliability of the product in order to achieve a balance. This balance will
subsequently trigger new orders and investments towards this direction and in
a long run a two Tier market, of upscale innovative ships and more
conventional ones, with the last struggling to face the competition both in
commercial and operating terms. It is the technology leap that will illustrate
the potential of the new designs and establish them as actual and realistic
solutions.
The present diploma thesis concludes with chapter 5 that describes a brief
summary and outline of the work undertaken, providing some design
directives for future, tender concepts as well as an outline of the Swath
optimization (global and less detailed one). The contribution of the Thesis in
Swath design is outlined and some perspectives for future work are
underlined.
Finally appendix A and appendix B contains two small case studies. The first
one is about an optimization of tri-hull with the same displacement of AQ3.
The main matter is to minimize the waves produced by the hulls behind the
main hull. The second one is an optimization of a tetra-hull with a diamond
shape again trying to minimize the waves of the hulls behind the main hull,
and hence the total resistance.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

6
1. CAD/CAE SYSTEMS FOR SHIP DESIGN

1.1Introduction

The tools and techniques used to design ship structures have evolved over
the last forty years, from producing blueprints on the drafting board to the
digital design of today. As computer technology became more powerful and
less expensive, computer-aided-design (CAD) systems evolved to support the
design of complex products. CAD and other related tools empower designers
and engineers to create innovative products more quickly and efficiently.

1.2 Objectives of CAD and CAE applied to hull forms

The ultimate objective of every tool used for economic human activity is to
obtain greater efficiency, effectiveness and a better quality. A greater
efficiency means that less time, material and labour are necessary to obtain
the desired results. Greater efficiency leads to:

 A shorter time to reach a certain design stage.


 Fast analytical calculations possible.
 Integration between CAD and CAE.
 Fast geometric manipulations.
 Increased job satisfaction.
 More freedom in the sequence of design activities (e.g. stability
calculationbased on a preliminary CAD model returns more accurate
information at the initial stage of the design).

To be of greater effectiveness implies that more topics can be dealt with,


which also lead to a better quality. For example:

 More design iterations, to come to an optimal design.


 Integration of analytical tools.
 3-D visualization, to give all persons involved a better image of the
vessel.
 Higher precision of the hull form definition.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

7
On the other hand, CAD and CAE systems bring also some disadvantages
with:

 The use of improper CAD/CAE systems, which force the designer into
a corner.
 The usual need for very powerful hardware system to support it;
 A tendency to use always the latest CAD/CAE products, which may be
unstable and error prone.
 A tendency to „over-calculate‟, just because the computer gives the
ability to, resulting in time-consuming procedures.

1.3 Ship design process

In ship design there are many domain-specific models of the design process,
but Evans‟ design spiral (below) is probably the most well known. This model
emphasizes that many design issues interact and must be considered in
sequence, in increased detail in each pass around the spiral, until a single
design that satisfies all constraints and balances all considerations is reached.
Modern CAD/CAE systems allow a holistic design approach which aims at
investigating many if not all important aspects at the same time. Such a
synthesis model of CAE (below) allows exploring the design space to a
greater extent and provides an efficient method of handling complex systems
with many relationships and dependencies at once.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

8
Figure 1.1 : Traditional design spiral[5]

Figure 1.2 : Integrated approach[5]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

9
1.4 An insight in fully-parametric ship design

In figure below the different modeling concepts are presented and compared
on the basis of flexibility, required knowledge, effectiveness and cost in
relevance to efficiency.

Figure 1.3 : Assessment of different geometric modeling techniques[14]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

10
What can be derived by the figure above is that the fully parametric modeling
technique yields excellent efficiency since only a few modifications are
required in order to achieve a new fair hull form. This approach requires
though a good knowledge of the basic elements of parametric modeling and
the most time is consumed in order to set up the whole structure. Once the
model is established, a wide variety of new designs is available, in contrast to
conventional modeling where, setting up a hull form and browsing through
new designs are equally time consuming and demands experience of the
designer. Partially-parametric models build on existing shapes and prove to
be an easy-handled approach for numerous tasks but it is not recommended
for global and multi-objective investigations since the allowed modifications of
the model are restricted. According to (Harries, 1998) , the great advantage of
parametric modeling is the ability to find the optimal balance between
variability and simplicity, more precisely the balance between the freedom to
be able to do everything and the restriction to do only what you really
need.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

11
1.5 FRIENDSHIP-FRAMEWORK

The FRIENDSHIP-FRAMEWORK is a CAE package for the design of


functional surfaces. It offers a wide range of CAD functionality for
conventional NURBS-modeling, partially parametric modeling with various
transformations and fully parametric modelling.
This software comes with a set of embedded variation and optimization
strategies. These algorithms can be comfortable linked to the geometry and
perform automatic variant creation. For that purpose, comprehensive variant
and constraint management are provided .
Any program or tool which is needed for geometry design and analysis can be
coupled. Convenient integration mechanisms make the external program an
inherent part of the FRIENDSHIP-FRAMEWORK. By doing so, design and
analysis expertise is centralized in order to streamline the design process.
CFD solvers are coupled to the CAD through various levels of integration;
tool- or project specific integration or by a common data interface. Therefore,
results of CFD computations can be easily used as measures of merit for
optimization procedures, driving the design process.
In addition to configuration and execution of external programs,
comprehensive post-processing functionality is available. Result data gets
visualized and tables are generated so that the entire design process finally
takes place within a single workbench.

1.5.1 Design principles

A typical design procedure within the FRIENDSHIP-FRAMEWORK starts with


a fully parametric model of the considered shape. During the geometry setup,
objects are related to each other via introducing dependencies. Changes that
are applied to one object are internally passed to dependent objects for
update purposes. Surfaces are no longer described via basic point data. More
intuitive descriptors (e.g. user-defined distributions which describe product
properties) help to modify geometry smartly in a way that the resulting
surfaces cover high fairness for geometrically feasible designs. Note that no
“black-box” models are used, the engineer is completely free to set up any
individual design. In the second step, parts of the geometry are linked
to variation engines. Any floating-point number of the model setup can be
varied. The user chooses a specific engine and defines bounds for variables
as well as constraints and objectives. In order to be able to assess the manual
or automatic variants, external software is coupled and configured. The
engines simply evaluate parameters that request an external value. This
transfers external data into the FRIENDSHIP-FRAMEWORK. Based on this
integration – along with parametric geometry variation – sophisticated formal
optimizations can be carried out .

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

12
1.5.2 Basic elements

The FRIENDSHIP-FRAMEWORK allows designing with a wide variety of


point, curve and surface types. Curve intersection point, NURBS curve, lofted
surface, Coons patch etc., are already known from other CAD programs and
are fully-functional. Within the FRIENDSHIP-FRAMEWORK there are some
special entities, which make the software a unique fully-parametric CAD tool.
The most important of these is the following:

1.5.2.1 MetaSurfaces

They are novel surface entities developed for collecting information


available in two distinct directions. They yield the Cartesian coordinates of any
point on the surface for any pair of surface coordinates u and v, basically
giving an unambiguous mapping from 2 to 3 as would, say, Bézier or B-Spline
surfaces, too. However, they are more flexible as they do not assume any
particular representation with regard to the curves they capture.

1.5.2.2 Feature modeling

Features are the way scripts can be implemented into a design. They work
the same way than libraries for a programming language.On one hand can be
produced with a little more than a click of the mouse instead of modeling them
from the scratch every time which would take quite a while but on the other
hand, the user has to be quite familiar with script writing, especially when
difficult geometries and concepts are required.

Some are already provided by the editor but for advanced object definitions
the user can create his own. They can for instance be used for interfacing the
FF with an hydrostatics software like Hydromax. The feature then requires a
meshed surface and a few data such as design waterline, displacement, LCG,
VCG, LCB and then calls Hydromax to perform the calculations. The features
then returns the results in the form of various objects, such as a righting lever
curve for instance that can be used in the design.

They can be created either as "persistent", in the case their result will be an
object incorporated to the current design, or transient if they are just needed
to perform a task on the design (a geometric transformation for instance).

Features encapsulate any user-defined command sequence and that makes it


available for writing macros and subroutines. They are high-level entities that
can offer readily shaped and parameterized elements, as opposed to primitive

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

13
elements like points, lines and "normal" curves and surfaces and represent
specific work processes which can be stored externally and reused.

Features work on the base of an editor where the necessary input parameters
and types are specified as arguments and then a process is described via
commands. Thins script is finally evaluated and returns the produced output
that makes up the feature‟s attribute. Features are flexible and can be
combined with each other providing sophisticated objects.

1.5.2.3 Curve engines

One of the useful objects for defining hull shapes in the FRIENDSHIP
FRAMEWORK is called the Curve Engine. It allows for a type of curve to be
defined at any position by a parameterization function of its position. In
opposition to a classic hull design software where a few sections are defined
at specific points along the hull length, the use of the curve engine allows to
have a continuous section envelope running along the whole hull:
Basically, the curve engine combines a "template" curve definition with a
continuous description of this definition. The template is formulated as
Feature definition and describes a curve by means of its configurable
parameters. Then, the curve engine takes this definition and connects the
parameters with a functional description of them, i.e. distribution. This means
that for each (abscissa) value in the interval of the functions, a curve can be
generated where the information stems from the input functions (i.e. the
ordinate value of them). (FRIENDSHIP Systems GmbH 2010)
Figure below shows a plot of a few sections generated by a curve engine. At
the bottom in yellow, the curve used to parameter the lowest position of the
section, that physically corresponds to the baseline. In black in the X-Z plane,
the curve used to parameter the beam of the sections along the hull. The
parameterization of a section is actually more complicated, and just two
template curves are shown for clarity purpose.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

14
Figure 1.4 : A plot of a few sections generated by a curve engine

1.5.2.4 Design engine

In order to fulfill this objective of performance driven design, FF has several


optimization algorithms implemented. The type of objects they belong to are
"Design Engines". These design engines can be used to either alter an
existing design or create new ones. These entities enclose several variation-
optimization algorithms, embedded in FRIENDSHIP-FRAMEWORK, which
are available for Design of Experiments, single-objective and multi-objective
optimizations. To name some: Sobol, Exhaustive Search NSGA-II, Mosa etc.
Design variables are chosen from the project which shall be involved in the
variation/optimization. For the most engines the lower and upper bound need
to be set, as well as the current value. Then, the evaluations are chosen,
which are parameters involved in the project. The results of the variant
creations are presented in a table, with the value of the variables used to
create them, the Fparameter object(s) on which the evaluation is running and
the constraints of the design that have been defined by the user.. The
evaluations can be set as objectives which then are minimized. Equality or
inequality constraints may also be involved. According to the underlying
algorithm, these constraints may be considered or not.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

15
2. Optimization
2.1 Introduction-The generic ship design optimization problem

Optimization is a process inextricably linked with human activity. The desire to


consciously optimize the outcome of decisions is a uniquely human character
trait .(Nowacki, 2003)
.
In a few words, optimization is the process of decision making when a number
of alternative choices are available and an optimal solution has to be
determined with regard to specific criteria, while taking into account the
restrictions and constraints set by the environment.

In more details, decision making of all kinds involves the choice of one or
more alternatives from a list of options. The list of options would normally all
be more or less acceptable solutions for the problem at hand and
consequences, both good and bad, flow from the exercise of choice. The aim
of rational decision making therefore, is to maximize the positive
consequences and minimize the negative ones. As these consequences are
directly related to the decision made or opinion set, it is not unreasonable to
treat the consequences as aspects of performance. The decision problem
then becomes a matter of considering these aspects of performance of all
the options available simultaneously so that the decision maker (DM) can
exercise his choice. In other words, rational decision making involves choice
within the context of multiple measures of performance or multiple criteria .

Ship design is a typical optimization problem involving multiple and frequently


contradictory objective functions and constraints [16]. With a system as
complex as a ship, composed of many subsystems that are complicated on
their own right, a naval architect is faced by a multiplicity of requirements,
from the owner‟s needs and desires, engineering feasibility, imperatives of
technological advancement, environmental considerations.

Solving the requirements of the sub-systems alone will often not produce an
ideal result; the interactions amongst the sub-systems must be analyzed,
leading to a ship design that truly is a multi-criteria decision problem. These
MCDM methods can vary in complexity depending on not only the amount of
parameters analyzed, but also how many of their interactions are thought out.
In addition, subjectively becomes a factor into determining which criteria stand
out above the others. How these criteria are weighted is up to the individual
method itself . Thus, the difficulty lies in formulating the objective and all the
constraints. For this reason, the main requirement when dealing with the
generic ship optimization problem is that the designer has a picture of his
objective, what he really wants to achieve .

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

16
2.2 Single Objective & Multi-Objective Optimization

In the classical optimization where there is only a single criterion and a set of
satisfiable constraints, decision making approaches lead naturally to the
solution. Once the criterion of interest is agreed upon -cost, for example- the
choice of the most attractive action is not a matter of opinion. There can be
some argument as to how the objective –cost, for example- may be computed
but in any meaningful problem the method of computation is obviously a part
of the definition of the criterion itself. In other words, the choice of criterion
leads directly to the solution in the mono-criterion example and it is a solution
that all parties can agree with.

However, sometimes a system must perform more than one mission or must
meet multiple objectives simultaneously or consecutively, which without
special assumptions may not be easy to accommodate in a single measure of
merit. Therefore, in the case of multiple criteria formulation, decision makers
can and will, in general, have different value systems leading to different
priority orderings of the multiple, potentially conflicting performance criteria .

Ten or fifteen years ago, standard available optimization tools would focus on
a single and limited aspect (e.g. shape, scantlings, propeller, ultimate
strength, etc.) and a single objective would be targeted (weight, resistance,
cavitation, etc.). Nowadays optimization tools tend to adopt a more generic
approach and coupled with the fact that they have also become much more
reliable this has made them more likely to be part of the standard design tool
set that each designer uses on a day to day basis.

So regarding to the multi-objective optimization we can understand that


requirements in design of any kind are often potentially in conflict. This is
because there are few, if any systems that can combine the best of all
performance aspects for all possible scenarios in the same design. If such
utopian solutions exist then the obvious answer would be to go for them. But
life being, the way it is, good values of some criteria inevitably go with poor
values of others. The aim in multiple criteria decision making is then to find
the best compromise solution . The Pareto optimality expresses exactly this
formulation, namely the Pareto optimal solution is a set of possible solutions,
a set of non-dominated solutions, in which no single objective can be
improved without degrading the achievement of at least one other objective.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

17
2.3 Generic optimization problem

The formulation of optimization problems is a conceptual modeling process


that follows certain standard procedures and results in a specific problem
definition, tailored for an application, e.g. in design .
From the viewpoint of information flows, the generic optimization problem and
its basic elements may be defined as follows (see Figure):

-Input EI: prescribed data, for example, requirements of the owner (DWT
capacity, service speed etc).
-Output EO: result of the evaluation of the system performance for given input
(techno-economical characteristics of the ship,- optimal solution based on
criterion/-a.
- Design variables D: free variables of the optimization problem (under the
designer‟s control), for example, ship‟s main dimensions.
- Design parameters P: restriction parameters, constraints (extraneous
influences, scenarios, side conditions, not under the designer‟s control).
- Merit functions M: measure of merit, expression of evaluation criterion/-a,
objective function (M(D,P)).
- Constraints G: boundary conditions of equality and/or inequality type,
function of design variables and parameters (G(D,P)).

Figure 2.1 : Optimization System[7]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

18
2.4 Solving a multi-objective optimization problem

As there usually exist multiple Pareto optimal solutions for multi-objective


optimization problems, what it means to solve such a problem is not as
straightforward as it is for a single objective optimization problem. Therefore,
different researchers have defined the term "solving a multi-objective
optimization problem" in various ways. This section summarizes some of them
and the contexts in which they are used. Many methods convert the original
problem with multiple objectives into a single objective optimization problem.
This is called a scalarized problem. If scalarization is done carefully, Pareto
optimality of the solutions obtained can be guaranteed.
Scalarizing a multi-objective optimization problem means formulating a single-
objective optimization problems such that optimal solutions to the single-
objective optimization problem are Pareto optimal solutions to the multi-
objective optimization problem. With different parameters for the scalarization,
different Pareto optimal solutions are produced. A well-known example is the
so-called linear scalarization :

where the weights of the objectives are the parameters of the


scalarization.

Solving a multi-objective optimization problem is sometimes understood as


approximating or computing all or a representative set of Pareto optimal
solutions.
When decision making is emphasized, the objective of solving a multi-
objective optimization problem is referred to supporting a decision maker in
finding the most preferred Pareto optimal solution according to his/her
preferences. The underlying assumption is that one solution to the problem
must be identified to be implemented in practice. Here, a human decision
maker (DM) plays an important role. (S)he is expected to be an expert in the
problem domain.
The most preferred solution can be found using different philosophies.
Generally multiobjective optimization methods are divided into four classes. In
so-called no preference methods, no decision maker is expected to be
available, but a neutral compromise solution is identified without preference
information. The other classes are so-called a priori, a posteriori and
interactive methods and they all involve preference information from the
decision maker in different ways.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

19
In a priori methods, preference information is first asked from the decision
maker and then a solution best satisfying these preferences is found. In a
posteriori methods, a representative set of Pareto optimal solutions is first
found and then the decision maker must choose one of them. In interactive
methods, the decision maker is allowed to iteratively search for the most
preferred solution. In each iteration of the interactive method, the decision
maker is shown Pareto optimal solution(s) and (s)he can tell how the
solution(s) could be improved. The information given by the decision maker is
then taken into account while generating new Pareto optimal solution(s) for
the decision maker to study in the next iteration. In this way, the decision
maker learns about the feasibility of his/her wishes and can concentrate on
solutions that are interesting to him/her. The decision maker may stop the
search whenever he/she wants to.

2.5 Design of experiment (DOE)

In general usage, design of experiments (DOE) or experimental design is the


design of any information-gathering exercises where variation is present,
whether under the full control of the experimenter or not.
In the design of experiments, the experimenter is often interested in the effect
of some process or intervention (the "treatment") on some objects (the
"experimental units"). Design of experiments is thus a discipline that has very
broad application across all the natural and social sciences and engineering.

So DoE‟s are very effective to gather information about the optimization


problem at hand and about the whole design space. DoE tables are useful to
detect trends of the optimization variables with regard to the objectives of the
problem. Alternatively, a DoE database may be searched to detect a suitable
starting point for a subsequent focused optimization process. Or a DoE may
serve as a database for response surface fitting, or for checking the response
sensitivity of a design candidate.

A design of experiment is used to identify which factors are statistically


significant and practically important to the overall design. Statistical
significance refers to the mathematical test to distinguish between whether a
design variable influences the change in the mean value of the outcome due
to an effect described in the model and whether the change could have been
observed in the data by chance alone. In essence, a design of experiment is a
research method that contributes to identify the changes, the local
inima/maxima, to get an idea about the shape of the objective functions and is
used as a preliminary tool for exploration of the design space and exploitation
of the best regions according to criteria in order to obtain a reasonable initial
design for the subsequent optimization.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

20
2.6 Genetic algorithms(GA)

In the computer science field of artificial intelligence, a genetic algorithm


(GA) is a search heuristic that mimics the process of natural evolution. This
heuristic (also sometimes called a metaheuristic) is routinely used to generate
useful solutions to optimization and search problems. Genetic algorithms
belong to the larger class of evolutionary algorithms(EA), which generate
solutions to optimization problems using techniques inspired by natural
evolution, such as inheritance, mutation, selection, and crossover.

Genetic algorithms (GA) are stochastic, nonlinear optimization methods that


apply the principles of biological evolution . In particular, they utilize
populations of solutions and apply selection, reproduction and mutation
methods, in contrast to more traditional optimization methods which use
gradient information to move between (successively better) points in solution
space. That makes them uniquely adaptive to multi-objective problems such
as finding Pareto frontiers.

A good definition provided by Koza (1998) is:

“The genetic algorithm is a highly parallel mathematical algorithm that


transforms a set (population) of individual mathematical objects (typically
fixed-length character strings patterned after chromosome strings), each with
an associated fitness value, into a new population (i.e. the next generation)
using operations patterned after the Darwinian principle of reproduction and
survival of the fittest and after naturally occurring genetic operations (notably
sexually re-combinations)”.

Actually, the genetic algorithm derives its behavior from a metaphor of one of
the mechanisms of evolution in nature which is called hard selection. Under
this scheme, only the best available individuals are retained for generating
descendants. This contrast with soft selection, which offers a probabilistic
mechanism for maintaining individuals to be parents of future progeny despite
possessing relatively poorer objective values.

A genetic algorithm for a particular problem must have the following five
components:

 A representation for potential solutions to the problem.


 A way to create an initial population of potential solutions.
 An evaluation function that plays the role of the environment, rating
 solutions in terms of their “fitness”.
 Genetic operators that alter the compositions of children.
 Values for various parameters that the genetic algorithm uses
(population size, probabilities of applying genetic operators, etc).

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

21
Some of the basic terminology referred to GA is the following:

The fitness of an individual is a value that reflects its performance (i.e. how
well solves a certain task). A fitness function is a mapping of the
chromosomes (data structure that holds a “string” of task parameters or
genes, analogous to the base-4 chromosomes present in our DNA) in a
population to their corresponding fitness values. A fitness landscape is the
hyper-surface obtained by applying the fitness function to every point in the
search space.

If the solution of a problem can be represented by a set of N real-values


parameters, then the job of finding this solution can be thought of as a search
in an H-dimensional space. This region is simply referred to as the search
space of the problem.

Exploitation is the process of using information gathered from previously


visited points in the search space to determine which places might be
profitable to visit next. Hill climbing is an example of exploitation, because it
investigates adjacent points in the search space, and moves in the direction
giving the greatest increase in fitness. Exploitation techniques are good at
finding local minima (or maxima). The GA uses crossover as an exploitation
mechanism.

Exploration is the process of visiting entirely new regions of search space, to


see if anything promising may be found there. Unlike exploitation, exploration
involves leaps into unknown regions. Random search is an example of
exploration. Problems which have many local minima (maxima) can
sometimes only be solves using explorations techniques such as random
search. The GA uses mutation as an exploration mechanism.

Elitism is a mechanism which ensures that the chromosomes of the highly fit
member(s) of the population are passed on to the next generation without
being altered.

The basic operation of a GA is presented in the following segment of pseudo-


code:

Generate initial population, G(0);


Evaluate G(0);
t:=0;
repeat
t:=t+1;
generate G(t) using G(t-1);
evaluate G(t);
until a solution is found

First, an initial population, where the individuals are set of chromosomes

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

22
representing all possible solutions to the problem, is randomly generated.
Then a fitness function is applied to each one of these chromosomes in order
to measure the quality of the solution encoded. Knowing each chromosome‟s
fitness, a selection process takes place to choose the individuals that will be
parents of the following generation.

Some brief explanations are now follow for the algorithms that were utilized at
the present diploma thesis:

Exhaustive Search:

Sobol:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

23
NSGA-II:

MOSA:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

24
3. MULTI HULL SHIPS AND SWATH
3.1 Multi hull ships

The growing demand for more space on the decks has led to the development
and improvement of double hull vessels, which are used both as speedboats
and as sailboats. The most common type of double hull vessels is the
catamaran, which consists of two hulls that meet the requirements for
buoyancy and stability and are linked by a superstructure, which gives the
total usable volume of the catamaran. Each of the two hulls called demihull
and may be symmetrical, asymmetrical or even fully asymmetrical to the
centerline.

The main advantage of the catamaran is the large surface area of the deck,
which makes it suitable for the carriage of passengers and vehicles. Note that
a catamaran vessel with respect to a monogastric vessel shows the same
displacement on average about 30% more surface deck. Additional
advantage of the catamaran is a fact that the total width is much larger than
that of a single-hull vessel, which is a particularly important factor for
economic exploitation. Noteworthy also feature of catamaran is not only good
behavior to the roll but also sufficient stability, partly due to the appropriate
choice of the distance between the two hulls and the large moments of inertia
of the two hulls, which contribute to the stability of the vessel and to reduce
the accelerations which lead to roll. Finally, as an advantage of the catamaran
can be mentioned the relatively lower draft in relation with a single hull boat of
the same displacement. As a result can float in shallower seas and harbors as
well as more freedom in selecting main dimensions that can lead to slender
hulls with little resistance, which is not always possible in the design of single-
hull vessels because stability requirements must be met.

The catamaran used since the early 1970s.By early 1990s started to
manufactured longer catamarans from 70 to 100 meters, with modern
tendencies to push catamaran more than 110 meters long and displacement
of 2000 tons. In general, the construction of catamaran ships is more
complicated than the construction of single-hull ships but simpler to
manufacture other speedboats. As construction material steel is prevented
and aluminum is preferred to the greatest possible reduction of the
displacement which is necessary to achieve high speed. Finally, for
propulsion water jets are commonly used and Diesel for main engines.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

25
3.2 General aspects of SWATH

The name Swath is an acronym that origins from the words: Small Waterplane
Area Twin Hull .

Ships of Swath type can be considered a special class catamaran. The


SWATH has all the advantages of catamaran vessels as large deck area and
comfortable satisfaction of the stability criteria.Their characteristics gives
them a lot of extra properties.

The SWATH consists of two lower hulls(gondolas) completely submerged


below the surface of the sea which offer almost the entire lift. Above sea level
looks like a common catamaran. The deck(box) links with the lower hulls
with thin bases called struts(these are usually two or sometimes four
ones).Struts are hydrodynamic designed in such way in order to reduce the
strong wave resistance for which they are primarily responsible.

Figure 3.1:Typical Swath hull [3]

The main feature that is understood by its name is the small waterplane area
that is accomplished from the immersion of the most of the displacement
below the sea level. Swath vessels has a waterplane area diminished about
50% in relation to a catamaran. The specifity of the hull offers extremely good
behavior to the sea waves. This is their great advantage in relation to the
other techonogicaly advanced ships which are forced to decrease their speed

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

26
in rough seas. Swath keep both their speed and the convenience of the
passengers.

The slender hulls have little interaction with the sea. Not only they stimulate
relatively small waves ,but also stimulation from the sea waves is reduced.
That is why they have reduced wave resistance and excellent handling in
rough seas.

The small wave resistance is small price relative to the increased friction
resistance due to the large wetted surface of the submerged lower hulls. So
for the low speed region where the frictional resistance is dominating the
efficiency of the hull is moderate. Contrary at high speeds where the wave
resistance is generally the major component of resistance, the Swath stands
excellent from hydrodynamic side.

Moreover, the small production of waves enables you to pass close to shore
without creating security problems in humans and boats.

A big problem although is the pitch instability espacially at high speeds. The
pressure field induced to the surface of the vessel movement generates a
tendency for increased trim (monk moment-nose diving).The solution used
today and has eliminated the problem is the placement of stabilizing fins
which are clearly shown in Figure 1.

3.3 SWATH Design Parameters

By changing some parameters can generate several alternative designs with


differing, in general, characteristics. These parameters are for a SWATH , the
waterline surface, the spacing of the hulls, the distance between the surface
of the sea and wetdeck etc. Each change causes positive and negative
effects. The challenge for the designer is to find the golden ratio.

The challenge in designing a successful SWATH vessel is not to minimize the


motion of the vessel at the expense of other characteristics such as the power
or the payload. Pursuing the movements of the ship to be significantly lower of

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

27
the commonly accepted limits for the comfort of passengers and crew. The
excessive decrease the water area plane has a number of unpleasant side
effects and problems such as lack of stability and high sensitivity changes in
weight and center of gravity.

The ship movement are caused by forces that the waves create on the struts.
These forces diminish with decreasing waterplane area of strut. With careful
selection of the waterplane area of strut is feasible to reset the vertical forces
due to wave to a narrow range of frequencies.

A decrease of the waterplane area makes necessary an increase the lateral


distance between the hulls to ensure satisfactory stability for the ship. Having
the hulls away from each other has a direct impact on the weight of the metal
structure of the ship which is growing rapidly with the unsupported length. So
typically reducing the water surface leads to a larger and more expensive
vessel.

A decrease to the size of the strut (especially to the beam) has a result of
space deficiency for the placement of the engine room and a lack of adequate
opening for maintenance. Moreover the construction is more complex. .
Examination of above parameters puts the problem in perspective and
highlights the need for deep understanding of how the Swath hull reacts,
before the optimization process.

Another aspect that deserves particular study is the gap between the sea
level and the bottom of the cross structure connecting the hulls together.
Reducing the gap essentially eliminates the big advantage of SWATH to
maintain service speed even in rough seas. From the other side the existence
of a large gap translates to a great height car deck making it difficult to
disembark and embark passengers and vehicles.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

28
3.4 Comparison of SWATH with conventional ships

Advantages of SWATH ships over conventional ships with the same payload
and relatively lower speed are as follows:

-Less required horsepower, especially in high speed areas.


-Superior seakeeping behavior
-Adequate stability in all conditions
-Extremely good maneuvering

The consequences of the above is important. First, relatively higher speeds


for the same power reduce the journey time and increase the number of trips
per year with many economic benefits for the owner, especially in peak
periods. The duration of the journey shrinks even more due to the reduced
time of loading and unloading of the ship in ports. As we know these ships
have a wider catapult. Moreover the excellent behavior of ships at sea, the
comfort due to the large deck space, low noise level and vibrations, aesthetics
and new technology that characterize this kind of ship are factors that make it
an attractive option for the passengers.

To the weaknesses of the vessel are:

- Large frictional resistance of the ship especially at low speeds,

-The sensitivity to changes in the weight and the center of gravity and finally

-The necessity of stabilizing fins to reduce pitching.

All these suggests that the ship is not suitable for large loads transfer.
Moreover, a Swath construction is expected to be more expensive than a
conventional ship.

3.5 History of Swath

Swath ships are not a new invention, although are considered by most
modern ship types and incorporate many new technologies. Below is a brief
history of the invention and evolution of this type of ship:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

29
1880 - 0 C.G. Lyidborg claims the patent for the first one-hull semi submerged
ship.

Figure 3.2 [3]

1938 - Appears for the first time the idea of a small-waterplane area twin huII
ship. The idea belongs to the Canadian Frederick G. Creed. Creed suggested
this type of ship as the most suitable aircraft carrier for the British navy. After
several years remade his proposal to the U.S. Navy but neither there was
echoed. In 1946 Creed enshrines his patent in Britain.

Figure 3.3 [3]

1959 - The U.S. Navy orients research on high-speed semi submerged ships.
In these investigations H. Boericke propose a one-hull ship called `shark
form'. In 1962 enshrines that invention.

Figure 3.4 Shark Form [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

30
1965 – Alan McClure from Houston suggests the ship “MOHOLE” for drilling
platform.

1967- Dr Reuven Leopold from Litton Industries suggests to the US Navy ths
craft “TRISEC”

Figure 3.5 TRISEC [3]

1968 - The first low speed Swath “DUPLUS” from the shipyard Boele in
Holland. The ship has a length of 40m and 1200t displacement. This is a
research boat for finding oil. The designer J. J. Stenger was based on the
observation that the submarines operating in small depth are influenced very
little by the surface wave.

Figure 3.6 DUPLUS [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

31
1968 - "A student of MIT suggests an optimized version of “MOHOLE” 'within
a course. Then some experiments are made in the university tank with the
model ship. He calls his vessel semi-submerged catamaran.

1968 - Dr. Tom Lang from Naval Underseas Center (NUC) of San Diego
evolves a design for “high speed with semi-submerged hulls”. Significant
development of the project is the addition of moving fins to stabilize the ship at
high speed. He puts the fins aft of thecenter of gravity of the vessel to control
trim and pitching.

Figure 3.7 [3]

1970 - Mitsui Engineering & Shipbuilding Co. from Tokuo initiates a research
to “semi submerged catamaran (SSC)”.

1972 - The construction of Swath “SSP KAIMALINO” on behalf of NUC. The


ship then arises after 13 months of research on NUC and 2.5 years of model
tests and analytical design. The ship is finally working in March 1973.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

32
Figure 3.8 SSP KAIMALINO [3]

1973 – Swath acronym is officially used by the US Navy.

1979 - Shipyards Mitsui Engineering & Shipbuilding complete the world's first
Swath intended for commercial use. It is the passenger ships “SEAGULL” with
capacity of 446 passengers,at 26.5 knots speed.

1991 - The first of a 4 Swath series is delivered to Military Sealift Command.


The vessel is designed by the U.S. Navy. We talk about “LONG VICTORIUS”
with 71,5m length.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

33
Figure 3.9 LONG VICTORIUS [3]

1992 - Shipyards Finyards deliver the first Swath cruise ship, the “Radisson
Diamond” of “Diamond Cruise Ltd” company. The ship has a length of 131 m.

Figure 3.10 RADISSON DIAMOND [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

34
1993 - The existence of the world's first stealth ship is revealed. This is the
sea shadow A-frame Swath with 50m length. The ship belongs to the USA
and is manufactured by Lockheed Missiles and Aerospace Co.

Figure 3.11 SEA SHADOW A-FRAME SWATH [3]

It is estimated that currently around 50 Swath ships are traveling in the seas
around the world. Their sizes range from very small vessels with a
displacement of less than 100 t, to the largest Swath ever built, the cruise ship
Radisson Diamond built in Finland in 1992, with 11,500 t displacement and
131 m length. Most of Swath ships are built in the U.S. and serve research or
military purposes. The material most preferred in the manufacture of Swath is
aluminum. The diagram below shows the distribution of ships in number
compared to their displacement.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

35
Figure 3.12 Size distribution of existing SWATH ships [3]

3.6 New-built Swath

Occasionally extensive research has been done on these ships and have
recorded some very innovative designs that exploit the features and
advantages of this type of hull. These are some of them:

FS Class 751 research vessel: Thyssen Nordseewerke recently studied and


built an innovative and contemporary research SWATH 3500 t. displacement.
The vessel has a diesel engine propulsion power of 4160 KW. It can travel
5000nm in its service speed of 15Kn. The choice of this type of hull is due to
the excellent performance of this vessel in waves. Small responses in relation
to the situation of the sea greatly helps the investigations this ship serves

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

36
already. Moreover, the large deck space is necessary for scientific equipment
and research needs. The ship was launched in October 2003 .

Figure 3.13 FS Class 751 [3]

Figure 3.14 FS Class 751 [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

37
AIiSWATH: Another revolutionary design effort of Swath ship is developed by
the famous Italian shipyard Rodriquez. Shipyards Rodriquez specialize in
designing high-speed passenger crafts and have a large and experienced
design department. In their effort to renew their proposals have turned to
Swath technology and have already come up with a design they call '
AIiSWATH '. 'AIiSWATH‟ designed to satisfy modern requirements such as
high service speed with low fuel consumption, little environmental pollution
and making reduced waves during the trip.

'AIiSWATH' with length of 70 m, will has less power than conventional vessels
in order to achieve the same speed. Reduced consumption is large
commercial advantage of this vessel especially in these times that fuel prices
vary in particularly high levels.

Moreover, the design focuses on reducing its output waves. This is a large
problem of high-speed ships and often become dangerous to the nearby
coast. There are many examples of speedboats which are required to reduce
speed when passing from the near shore due to the large making waves
which is dangerous for humans and small boats. Speed reduction moderates
the most important advantages of high speed ships, their speed.

Experiments with AIiSWATH model were very encouraging so at October


2004 started to built in the shipyards of the company the first ship of this type.
The ship has a capacity of 500 passengers, 50 cars and a displacement of
only 500t. The company believes that this design would be the future to the
high-speed passenger ships.

Figure 3.15 Vessel ALISWATH [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

38
Damen Royal Schelde yard: Two new passenger Swath manufactured in
Damen Royal Schelde yard. Manufacturer justifies the choice of SWATH
saying they are in his priorities seaworthiness and reliability of ships. Swath
ships renowned for their seaworthiness even in rough weather conditions
where other vessels cannot make it. For reliability purposes was chosen
diesel-electric propulsion that is also characterized by high flexibility. The ship
can travel satisfactory even in case of loss of one of the two diesel engines
that propel the main generators. The two vessels operate between Vlissingen
and Breskens and put into the sea in March 2004 .

Figure 3.16 Vessel PLANET [3]

Explorer: Minesweeper SWATH with a length of 25 m from Aberking &


Rasmussen shipyard in Germany. The 25 meter long Swath is used by the
German navy. It is operating in the North Sea since 2004 after successfully
passed strict tests on the behavior at sea waves because it has sensitive
radio control equipment for demining .

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

39
HSSS 1500 – HSSS 900: These two Hybrid - semi SWATH serve routes
between Europe and England with great success. They offer extremely high
speeds, excellent behavior in waves and are the finest examples of
commercial application of SWATH.

Table 3.1 Characteristics of HSSS 1500- STENA EXPLORER and HSSS 900- STENA [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

40
Finally we should also mention the highly successful and internationally
recognized Swath designs made in the past from the”ship‟s design” laboratory
of EMP. It is about the passenger „Aegean Queen‟, the research vessel
`SMURV ' and the speedboat containership ‟SMUCC'. Ships have the
following characteristics:

Table 3.2 Characteristics of Aegean Queen, SMUCC FEEDER and SMURV [3]

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

41
4. MICHLET

Before MICHLET is presented and explained the way it works it is wise to


mention some things about components of resistance and thin ship theory.

4.1 Components of ship’s resistance

When a ship is moving through water there will be forces opposing the
motion. The total resistance, Rt, of a ship is defined as the force needed to
tow the ship at a constant speed and it can be divided into subcomponents in
different ways (Figure below).

Figure 4.1 :Components of resistance [14]

One way is to divide it into skin friction resistance RFO, and residuary
resistance RR, which includes all components related to the three dimensional
form of the ship and wave-making resistance. It can also be divided according
to physical phenomena into viscous resistance, RV, and wave resistance, RW .

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

42
Further on, the viscous resistance consists of the frictional and pressure
component. So the elaborated subdivision looks as follow:

The total resistance of a ship can be divided into three main parts [10]:
 wave resistance
 frictional resistance
 viscous pressure resistance

For each of the parts of the total resistance different effects are primarily
causative. Wave resistance depends on the lost energy due to the wave
production of the ship as a partially submerged body disturbing the free
surface of a fluid, thus waves are created due to water particles being
removed from their equilibrium position. Secondly, sheer stresses between
parts of the fluid with different velocities are the
reason for the frictional resistance. These sheer stresses occur in the area
close to the wall, within the boundary layer. Directly at the surface of the body,
or at a wall, the fluids velocity is equal to zero, but at the outer end of the
boundary layer the velocity is equal 99% of the undisturbed fluid velocity.
Viscous pressure resistance consists of effects like flow separation and
turbulence, which are mainly appearing in areas where the velocity of the fluid
is decreasing and therewith the thickness of the boundary layers is increasing.

4.2 Thin ship theory

This section was taken verbatim from E.O. Tuck (1987).


The thin ship theory of Michell represents the ship by a centre-plane source
distribution proportional to its longitudinal rate of change of thickness (local
beam). The only requirement for its validity is that that quantity be small.
Hence the theory applies as well to submerged as surface piercing bodies.

Another way of putting this argument is that slender bodies are also thin. To a
certain extent, slender body theory is superfluous. Even where a body looks
more slender than thin, it is unlikely that (to leading order) modeling it by a
centre-line singularity achieves more than modeling it by a centre-plane
singularity.

Indeed, there are examples where one can prove that this is the case, and
wave resistance is one of them. Thus, for surface-piercing ships, Vossers
(1962) and I (Tuck 1963) derived the slender ship wave resistance formula
laboriously from first principles, but Maruo (1962) was able to show very
simply that it could be obtained from Michell's thin-ship formula by letting the
draft go to zero.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

43
So why use slender body theory at all. This is a good question to which there
could be several answers. One is to dispute the evidence that thin ship theory
subsumes slender body theory. Since there is no exact formula for wave
resistance, one cannot say from the theoretical point of view which is the
better approximation. So the slender body result (even though it is an
“approximation” to the Michell result) could perhaps be more
accurate as an estimate of the true wave resistance. Many doubt this very
much. There are elements in the Michell theory (such as exponential decay
with depth) that are absent in the slender body theory, and it would be better if
would be present – although some people might assert the opposite. Anyway,
the evidence of comparison with experiment seems to favour Michell.

The thin and slender body theories only agree to leading order in slenderness.
If one expects to improve upon either theory by including second-order
contributions, there is a case for picking the more appropriate model, but
there have been very few attempts at second-order theories, and no
successes as far it is known. This is certainly irrelevant to the present review,
since nothing second order has been proposed for SWATHs, or should be.

The remaining argument for slender body theory is that it is a valuable


computational simplification. That is, if the body is indeed slender as well as
thin, we might as well make use of that fact in some way. Certainly there is
one immediate simplification in the data preparation, since the shape of cross
section is irrelevant in slender body theory, but input to thin ship theory. It is
believed this is such a minor matter, hence will be ignored. I feel that the
composite “slender body and thin strut” approach adopted by “Salvesen et al”
is hardly less difficult to implement than a full thin-body approach, and, when
the third (controversial) body-strut source distribution is also put in, any
advantage of simplicity disappears.

4.3 The code MICHLET

This section was taken verbatim from the Michlet manual.


MICHLET is a computer workbench, that was created by Mr. L. Lazauskas,
and can be used for investigations into some aspects of ship hydrodynamics.
Although it is not a ship design program, MICHLET can be used for
preliminary design work such as estimating the resistance, wave elevation
patterns and bottom pressure signatures of mono-hulls, multi-hulls and
submarines.

MICHLET calculates the total (viscous + wave) resistance and far-field wave
elevations of thin monohulls and multihulls using the ship thin theory.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

44
MICHLET is not an easy program for newcomers to hull design. A familiarity
with the first few chapters of an introductory Naval Architecture text would be
useful. There are several good WWW articles that might serve as sufficient
introduction.

There are many opportunities to waste a lot of computer time and human
effort with MICHLET. Incorrect specification of parameters in the input file
could result in completely useless output and it is imperative that users gain
familiarity with the input file and its nuances.

4.4 Basic considerations

This section was taken verbatim from the Michlet manual.


Co-ordinate System:

Figure 4.2: Co-ordinate system for general multihulls[22]

The co-ordinate system and hull parameters are shown in Figure 1.X is
positive astern, y is positive to starboard, and z is positive upwards. The
undisturbed water surface is the plane z = 0. Hull centre-planes are parallel to
the x-axis. The shapes of the hulls can be different from each other, however,
they are assumed to be symmetric with respect to their own centre-planes.
The overall vessel need not to be laterally symmetric.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

45
Each hull has a nominal centre point (x = si, y = wi, z = 0). The centre of the
main hull, hull 1, is always located at the origin. The centre of all other hulls
are then measured relative to the main hull, i.e. to hull 1 s i is the longitudinal
separation distance of the ith hull measured relative to the first hull and is
defined as positive astern. Hence by definition, s1 = 0. Wi is the lateral
separation distance of the ith hull relative to the first hull (positive to starboard).
Hence by definition, w1 = 0. For example, in the arrangement shown in
Figure1, s2 has a negative value, and w2 is positive.For hull 3, s3 is negative
and w3 is positive.

4.5 Calculation of resistance

This section was taken verbatim from the Michlet manual.


MICHLET calculates a variety of resistance components. Wave resistance is
estimated using Michell's thin ship theory augmented for (among others)
transom stern effects and boundary layer displacement thickness
effects.
For mono-hulls and for the individual hulls of a multihull, output includes:
Rh : Hydrostatic resistance on transom stern
Rf : Skin friction
Rwtrans : Wave resistance of transverse wave system
Rwdiv : Wave resistance of diverging wave system
Rw = Rwtrans + Rwdiv
Rform : Form drag
Rt : Total resistance
Rr : Residual resistance = Rt – Rf

Note that for supercritical depth-based Froude numbers, Rwtrans is always


equal to zero: there are no transverse waves!
As well as the components defined above, MICHLET also calculates the
following interference resistance components:
Rwtinter : Interference resistance of transverse waves
Rwdinter : Interference resistance of diverging waves
Rwinter = Rwtinter + Rwdinter

Note that for these interference components, positive values mean no


beneficial interference, (i.e. resistance is increased) negative values mean
that there is some beneficial wave cancellation (i.e. resistance is decreased).

4.6 Input files

This section was taken verbatim from the Michlet manual.


MICHLET uses plain text files for input and output. In addition, some plots are
saved as pcx graphic files. Text files can be read and saved using plain text

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

46
editors. Please note that you cannot use word-processors (for example,
Word) to edit MICHLET input files, unless you force them to save files in
ASCII format.
There are many shareware graphics programs that can be used to view pcx
files and to convert them to other graphic formats if necessary.

When MICHLET is run, it reads in a file named in.mlt and clears all output files
in the directory where the MICHLET executable files reside. If you do not save
previous output, old results will be overwritten and lost.

in.mlt

Initial input to MICHLET is via the plain text file called in.mlt.
At this stage, it is probably a good idea to have an example in.mlt file loaded
into your text editor, or to have a printout of one handy.
Comments can be placed in the file by preceding them with the # symbol,
which should appear in the first column of the line. Comments should be no
longer than 79 characters, and should not be placed within a column of
numbers.
.
If errors are encountered while the in.mlt file is being read, a message will be
displayed on-screen and also written to the out.mlt file before the program
terminates. Tracking down errors in the in.mlt file can be a little tricky.
Sometimes an error in one input line will cause an error to be reported for a
line further on in the in.mlt file. If an error is encountered, and no immediate
reason for the error can be discerned, check a few lines back in the file to see
if something was improperly specified.

4.7 General input files

At this point it is necessary to make clear that this section was taken again
verbatim from the Michlet manual.
The Input File Type and Input File Subtype are used to control the types of
input file that can be used. Similarly, Output File Type and Output File
Subtype are used to control the types of output files that are generated.

Course Particulars
The Course Particulars parameter is reserved for future enhancements.

Number of Vessels
The number of hulls for the vessel or ensemble must be an integer equal to 1,
2, 3, 4 or 5.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

47
Fluid Properties
The gravitational acceleration (in ms-2) should be entered as a decimal. Most
MICHLET examples use a value of g = 9.80665 ms-2.
Water density (in kgm-3) must be entered as a decimal. Most example files
use a value of 1025.9 kgm-3, the density of sea water at 15o, or 999.0 kgm-3,
the density of fresh water at 15o. Water kinematic viscosity (in m2s-1 x 10-6)
must be entered as a decimal. Most example files use a value of 1.18831 (sea
water at 15o), or 1.13902 (fresh water at 15o ). Note the units that are used.
The (non-dimensional) base eddy kinematic viscosity, vB, must be entered as
a decimal. Most example files use a value of 10.0. This quantity, which is not
a property of water alone, depends on the nature of a particular flow.
The main effect of eddy viscosity is to damp high frequency waves. However,
if a wave pattern seems to be corrupted by spurious, very high frequency
waves, it could be because the value for Nθ is too small. First try increasing
the value of Nθ then, if that doesn't fix the problem (and if it is actually
a problem) increase the size of the base eddy viscosity.
The eddy kinematic viscosity, vT , is calculated according to the formula:

vT = v + LUvB (1)

where L is the length of the vessel, U is the ship speed, and v is the kinematic
viscosity of water.
Water depth (in metres) must be entered as a decimal. To simulate infinite
depth use a large value (e.g. 10000.0m). You must ensure that hulls (in their
squatted attitudes) do not run aground.The next four parameters are not used
in any calculations in this version, however the program expects appropriate
values to be present.
Air density (in kgm-3) must be entered as a decimal. Most example files use a
value of 1.26 kgm-3. Air kinematic viscosity (in m2s-1 x 10-6) must be entered
as a decimal. Most example files use a value of 14.4.
Wind speed (in ms-1) must be entered as a decimal.
Wind direction (degrees) must be entered as a decimal. The convention is
that 0o corresponds to a head wind; 180o signifies a tail wind.

Ship Speed Range


The next three parameters specify the speeds (in ms-1) at which to calculate
the resistance and wave elevations, via the minimum speed, Umin, the
maximum speed, Umax, and the number of speeds, NU.
The minimum speed and maximum speed must be decimals. NU must be an
integer greater than or equal to 2 and less than or equal to 101.

Leeway
The Leeway parameter is reserved for future enhancements.

Number of Stations and Waterlines


The number of stations, Nx, used to represent the hull surface must be an odd
integer greater than or equal to 5 and less than or equal to 81. The number of

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

48
waterlines, Nz, must be an odd integer greater than or equal to 5 and less
than or equal to 81. The actual bow and stern ends are counted as stations.

Ship Loading Type


The Ship Loading Type determines the formula to be used in the calculation
of the distance from the baseline to the centre of gravity of the ship (i.e KG, in
naval architectural parlance). In this version of MICHLET, the only allowable
value is 3, which requires three comma-separated Ship Loading Formula
Parameters to be specified on a separate line. Suppose that these three
parameters are a,b and c. The three-parameter formula used in this version of
MICHLET is:

KGOA = a(TOA + bLOAc) (2)

Where TOA is the maximum draft, and LOA is the overall length of the ship.
For monohulls TOA and LOA are just the draft and length of the hull,
respectively. If you are unsure of where the centre of gravity is located for
your vessel, a reasonable rough estimate is the maximum draft. In this case,
use the following line in the input file : 1.0,0.0,0.0
Remember to take into account the fact that you have used a rough estimate
when interpreting your results.

Wave Resistance Parameters


Nθ is used in the calculation of the wave integrals, e.g. see Tuck and
Lazauskas . It must be an even integer greater than or equal to 10 and less
than or equal to 4096. For reasonable accuracy (2-3 figures) of wave
resistance calculations Nθ should be set to at least 160 for mono-hulls, 320 for
double-hulls, 640 for tri-hulls, 960 for tetra-hulls, and 1280 for penta-hulls. The
value should be set to larger values for wave elevation calculations far from
the ship. Note that some example files distributed with MICHLET use smaller
values than these recommendations. The purpose of the examples is to
provide quick demonstrations of certain features. Accuracy is of secondary
importance.
Before embarking on your own design problem, check that N θ is large enough
to ensure accurate solutions, but not too large so that run times are
unbearably long.
Before calculating wave resistance and other ship wave quantities, MICHLET
first adds an estimate of the boundary layer displacement thickness to the hull
offsets. In general, the effect of the boundary layer is small, except for small
model-sized hulls. For larger ships travelling at high speeds the effect is
negligible.
The default in MICHLET is to include boundary layer effects, and these can
increase execution time. To perform calculations without boundary layer
effects use a negative value for Nθ. For example, if the value

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

49
-160 is input, this will cause MICHLET to use 160 intervals of θ in all wave
calculations and the boundary layer displacement thickness will not be added
to the hull offsets.

Viscous Resistance

Skin-friction

Choices for the skin-friction method are:


0 : None
1 : ITTC 1957
2 : Grigson's algorithm

Form factors

Form factors can be applied separately to the viscous drag component and to
the wave drag component.
Choices for the viscous form factor method are:
0 : None
3 : Dual
The viscous drag form factor can be useful for analyzing (and designing for)
the effects of hull fouling. For example, if we assume (by rule of thumb) that
there is a 0.1% increase in friction resistance per day, then after 6 months we
should make an allowance of approximately 18% beyond the usual ITTC
friction. In this case, use a value of 1.18 for the viscous form factor.
It is important to note that the ITTC line is not a skin friction line. The ITTC line
is considered to be a correlation line and as such it contains some allowance
for three dimensional (i.e. form) effects. If further form effects are
included, there is a danger of double-counting. In some cases, (e.g.
correlating experiments and computer estimates) this will not necessarily be a
problem. The dual viscous drag form factor can be used when you want to
apply a form factor to the wave resistance as well as to the skin-friction.

Sea-Bed Pressure Signatures


Pressures exerted by the vessel onto the bottom of the ocean can stir up
sediments and trigger some types
of anti-ship mines.
Choices for the pressure signature method are:
0 : None
1 : Slender body approximation due to E.O. Tuck. Valid only for slender
vessels and for sub-critical depth-based Froude numbers (i.e. Fh < 1).

Wave Pattern Domains


Wave elevations can be calculated over two differently-shaped domains.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

50
In Figure 2, the co-ordinate origin is at the centre of Hull 1 (refer back to
Figure 1). The sectorial patch on the left of Figure 9 requires five parameters.
R0 and R1 (both decimals greater than 0.0) determine the (radial) extent of
the wave elevation patch. The value of β determines the angular
extent of the patch either side of the track of hull 1. The number of radial
nodes, NR and the number of beta nodes, Nβ, (both integers between 2 and
200, inclusive) determine the fineness of the grid for the calculation of wave
elevations. A value of 100 for both parameters gives reasonable results for
reasonable running times.

Figure 4.3: Wave elevation contour domains[22]

Hull Offset Data

It does not matter whether the offsets describing each hull are in dimensional
or non-dimensional form. MICHLET will automatically scale the offsets to the
individual hull displacements. It is very important to note that all hull offsets
input to MICHLET, or output by MICHLET, are for the underwater portion of
the hull only. You can, however, specify a value for the sinkage to raise or
lower
the hull with respect to the undisturbed free surface.

All offsets at the bow (the first row) must be equal to zero (decimal). Stern
offsets (the last row) may all be zero (no transom) or some non-zero if there is
a transom stern of a shape determined by the non-zero offsets. The number
of rows (cross-sections) and columns (waterlines) in the offset data must be
the same as the number of stations and waterlines specified earlier. Offsets
are separated by commas, and there is no comma at the end of each row.
Hull offsets in comma-separated format can also be read from file. For
example, if the input line for the first hull contains only the value -1, then
MICHLET will read offsets from the file useroff1.csv. If the offset input line for

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

51
the second hull was specified in the same way, MICHLET would read offsets
from the file useroff2.csv. Similarly for hulls 3, 4 and 5.

Hull Geometry

Displacement volume, hull length and draft


Following the offset table for the first hull, the displacement volume, length
and draft must be specified.

Demihull separation distances


After the draft has been specified, the longitudinal and lateral separation
distances (relative to the first hull) must be specified. For a monohull, these
will both be set to 0.0, irrespective of the values entered by the user (refer to
Figure 1).

Loading type
The Loading Type determines the formula to be used in the calculation of the
distance from the baseline to the centre of gravity. The method to be used is
identical to that used for the Ship Loading Type. Here though, the loading
applies to the hulls individually. For a multihull, the individual hulls can be
loaded using one formula, and the KG of the ensemble as a whole can be
calculated using another formula. This can lead to interesting results during
the optimization of multi-hulls.

Trim and sinkage


Trim must be specified in degrees (positive bow
up) and must lie between -10o and 10o. Sinkage is positive when draft is
increased, negative when the hull rises out of the water.

Appendages
The Number of Appendages should be left as 0 in the present version of
MICHLET.

Other Particulars
The Other Particulars field is reserved for special applications and should be
left as 0 in this version of MICHLET.

Multihull Geometry
If the number of hulls was set as 1 earlier in the in.mlt file, any further lines in
the input file will not be read. If the number of hulls is 2 or more, then the
details of the second hull, and the displacement volume, hull length, and hull
draft, etc of the second hull are also required. Similarly, for the third, fourth
and fifth hulls.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

52
5. CASE STUDY – DESIGN AND OPTIMIZATION OF AQ3
5.1 DESIGN OF AQ3
5.1.1 Parametric model design

The design of such a vessel is based on economical and technical demands


of the ship owner , physical restrictions , regulations about functionality and
safety the new technological evolutions. The engineer must couple all the
above to bring a result that is economical optimized with a productive exploit
of the initial fund.
This coupling becomes even more difficult when the vessel is a relatively new
type and basic elements are not easily provided. The design procedure
though is again the known helicoidal curve. Since it is a parametric model
FRIENDSHIP FRAMEWORK is one of the best software that you can work
with. Something that needs to be mentioned here is that the designing
procedure consists of many steps. In order to come up with an innovative
model these steps must be included at a loop so all imperfections should be
eliminated.
Now at the beginning a parametric conventional model was created within
the GUI interface of FRAMEWORK. This was composed by three parts.

GONDOLA
The first part was the gondola. Gondola is the lower hull of the ship that holds
most of the displacement and it is fully submerged under the free surface of
the sea. It is assumed that the vessel has a standard form with the following
characteristics:
-Lower hulls are of circular cross section.
-Lower hulls consist of ellipsoidal nose, parallel midbody and parabolic tail
whose lengths can be varied.
-Lower hull axis are horizontal parallel and symmetric to the centerplane.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

53
Figure 5.1 Gondola

STRUT
The second part is the strut. This connects gondola with the deck. Struts must
be hydrodynamic optimized so that the wave resistance must be as less as
possible. Again some assumptions must be made and they are as follows:
-Struts lie in vertical planes , single strut per hull concept.
-Strut profiles parabolic nose ,parallel midbody and parabolic tail. Their
lengths can be varied but in a standard rate of gondolas lengths.

Figure 5.2 Strut

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

54
BOX
The third part is the box. It is the deck of the vessel where passengers are
accommodated. It was not studied in depth because it does not offer to the
hydrodynamic resistance of the ship, therefore it was not optimized and
design only for aesthetic purposes.

Figure 5.3 Swath(with box)

5.1.2 INITIAL PARAMETERS


So on the basis of these geometrical characteristics the wetted part of the
swath part can be described in the terms of the following parameters:

 Lengthgondola= lower hull length


 Lgn=lower hull nose length
 Lgt= lower hull tail length
 Rad= lower hull radious
 Draft= lower hull draft
 Latsep= half distance between lower hulls
 Bstrut= Strut beam
 Lstr_n=strut nose length
 Lstr_t=strut tail length

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

55
So this was the initial (conventional as referred above) model. With that model
we could make a global optimization. But it was missing a serious parameter
that was not included so that we could not address a local form optimization.
A parameter that could express the displacement distribution. That was the
reason that the previous model was not sufficient. So a new model had to be
created that could include this parameter. In order to include the
aforementioned parameter feature programming was a must. FRIENDSHIP
FRAMEWORK has a kind of surface that called metasurface. This is created
by a section and a guide curve. Firstly the section is created then the curve
and finally the metasurface is created as the section is guided by the guide
curve. With this way we do not have a parallel midbody as we mentioned
before and we have a curve that when we change it, the displacement
distribution can be varied. The name of the last parameter is radmid and we
include it at optimization.

Figure 5.4 Model without a parallel midbody

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

56
Figure 5.4 Model without a parallel midbody

In this point we have to mention that from the above parameters we have to
remove Rad from independent and to make it dependent. This is because
when of one or more parameters varies, the displacement changes. But we
need a constant displacement of 1000t. So we have to express the volume
analytically and find the Rad that leads to the desired displacement. This
procedure must be done automatically when any of the parameters changes ,
so a algorithm is created within the FRIENDSHIP FRAMEWORK to do this
job.

As for the Draft parameter in order to validate the results of Aegean Queen
we had to keep it constant too at 5m. The only constrain was the following:
At the operation of the vessel displacement changes are expected to be
made. To be at the safe side an extreme case was considered so that if we do
not face problems at it we will be confident that no problems should be turn
up. So a 10 percent change to the initial displacement was considered and
the draft was checked. With the Awl existed there was no problem at the draft
that lied above 4m.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

57
5.1.3 FINAL PARAMETERS
After all the above again on the basis of these geometrical characteristics the
wetted part of the swath part can be described in the terms of the following
parameters:

 Lengthgondola= lower hull length


 Lgn=lower hull nose length
 Lgt= lower hull tail length
 Radmid= Displacement distribution parameter
 Latsep= half distance between lower hulls
 Bstrut= Strut beam
 Lstr_n=strut nose length
 Lstr_t=strut tail length

5.2 OPTIMISATION PROCEDURE

To investigate and develop innovative solutions, the designer requires a tool


that does not enforce detailed definition and allows easy reconfiguration of
arrangements and systems. Looking at the case of a innovative Swath, a CAE
environment(FRIENDSHIP FRAMEWORK) is established to examine key
measures of merit for a considerable numbers of variants simultaneously:
Geometry, capacity, and hydrodynamics were computed by means of
simulation codes.

In the case of this diploma thesis, the definition “optimization model” implies a
whole system that is built up by several subsystems in order to approach the
design of a novel Swath in a holistic way. The target of this holistic approach
is to create a fully parametric model able to vary in a wide range of
dimensions and form parameters and to return a large number of valuable
information predicting numerous features and properties of the subject.
Further on an optimization process takes place in several steps based on the
collected data, obtained from a large number of design variants.

Based on simulation driven design the designer is able to handle as many


issues as possible simultaneously. Obviously this is not a trivial task, therefore
a number of assumptions have to be made and the design problem has to be
viewed by specific perspectives, in order to define the boundaries and the
targets of the investigation. The main idea is to set up a large flexible model
that is able to predict automatically and accurately a large number of
properties regarding geometry, resistance and capacity. Therefore a number
of specialized tools have been integrated that allow a thorough insight and
increase the efficiency of the results. Thus, a numerous designs are achieved
quickly and accurately that enable comparisons and optimization.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

58
5.3 HYDROSTATICS

One of the main steps is the hydrostatics calculation. The calculation of the
hydrostatic quantities of every design is realized within the
FRIENDSHIP-FRAMEWORK based on an embedded computation for a given
configuration. The configuration receives as input the hull in form of offset
groups through a mechanism of automated adaption. Usually 3-4 offset-
groups are included regarding the complexity of the design. In our case only
one offset group was enough since the geometry is not so complex. But the
density of the offsets was changed regarding the complexity of its parts along
the vessel . Additionally, the sinkage and heel of the vessel are required input
as well. In this case no heel is considered.

When the configuration is ready the FHydroComputation is triggered which


returns a table with all basic hydrostatic quantities .
Further properties that derive from hydrostatic calculation such as coefficients
KM, BM etc are manually defined and are automatically updated for every
computation.
An illustration of SAC and of the submerged body in the form of section is also
given, pointing out the position of LCB and LCF. Another important quantity is
the estimation of the wetted surface, which is also needed later on when
MICHLET is called in the optimization problem. For this estimation an
appropriate feature, embedded in FRIENDSHIP-FRAMEWORK is used. This
feature executes the computation of the wetted surface for a given section
group at specific draft. For this feature a section group of 90 sections that
derive from the hull body, is utilized, returning an accurate wetted surface
estimation. Last but not least is the waterplane area Awl which is a very
significant quantity for the integrity of the draft as mentioned before.In the
following pictures are presented some illustrations of the hydrostatic
properties and the SAC.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

59
Figure 5.5 Sections of single hull

Figure 5.6 Offset density

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

60
Figure 5.7 SAC

5.4 MICHLET INTEGRATION

After the parametric model was created and the hydrostatics configuration
and computation were embedded, a resistance code had to be inserted in
order retrieve evaluations. So MICHLET coupling had to be accomplished.

FRIENDSHIP FRAMEWORK has some connections available with external


software like SHIPFLOW Hydromax, Neptune etc. But unfortunately is does
not have an available connection with MICHLET. So a new generic
integration( or “new software connector” as it was renamed at the new version
of FRIENDSHIP FRAMEWORK) had to be set up. This integration is generally
not too complex. Just a definition, a configuration , a computation, and a
connector that analyses input files and geometry and returns the output
results. In this case though, there was a serious problem that had to be
solved. MICHLET is an executable that needs user interference in order to
extract the results. Some specific buttons must be pressed with a time delay
because of the intermediate calculations. So a batch mode of MICHLET was
needed. At a contact with the designer of MICHLET I was informed that there
was not a batch mode of this code available. So a script in visual basic was
created in order to overcome this difficulty. As a result a FRIENDSHIP-
MICHLET integration was feasible with the support of this visual basic script
inside a batch file. Some basic photos are illustrated here:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

61
Figure 5.8 Connector

Figure 5.9 In and out files processing by FRIENDSHIP

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

62
After the integration was achieved the input files were connected to the
parameters and the output files with the evaluations in order to automate the
procedure. The last step after all these was to give the geometry to MICHLET.
Back again to feature programming an algorithm was needed to be created
and to extract the geometry from FRIENDSHIP, whenever a little change is
made, to the MICHLET input files.

5.5 EVALUATIONS OF THE OPTIMIZATION

Throughout the investigation a basic structure regarding the optimization


process occurs in two stages as listed below:

-Automated and systematic exploration of the multidimensional design


space with analysis sequences for more than 500 designs.
- Automated deterministic detailed optimization of best selected range from
exploration, with multi-objective genetic algorithms, utilizing many form
parameters.

At the very beginning, the design engine called, Exhaustive search is used.
This is an algorithm used in designs of experiments, where objectives are only
evaluated. This method of identification of the problem is used only at the
early stage.

Later, in order to gain a better insight into the design space and obtain a
reasonable subsequent optimization, a design of experiment is set with a
SOBOL Design Engine, embedded in FRIENDSHIP FRAMEWORK.

The deterministic SOBOL algorithm is a so-called quasi-random or low


discrepancy sequence and imitates the behavior of a random sequence. It is
more efficient and less random than a (pseudo) random number sequence,
which spreads points randomly in the design space. These sequences use a
base of two to form successively finer uniform partitions of the unit interval,
and then reorder the coordinates in each dimension . In this way, a uniform
sampling of the design space is attained, offering a better overview of the
design space, depending on the density of available variants. Sobol type
algorithms are known to have superior convergence than random sequences .

For the multi-objective optimization, the Non-dominated Sorting Algorithm II


(NSGAII) is utilized. The main advantages of this algorithm are that it applies
Pareto-based ranking schemes and avoids “trapping” between local maxima
(or minima). More information about the NSGA-II may be found at the
theoretical session.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

63
The main objectives that are monitored and optimized during the optimization
procedure reflect clearly the scope of this project. The holistic design
approach is implemented to this design problem with the following objectives:

 Minimization of the total resistance(Rt) at V=15.5Kn

 Minimization of the total resistance(Rt) at V=31Kn

Moreover some constraints should be introduced in order to get feasible


designs. The main constraints used were:

 Displacement evaluation should not diverge from 1000t

 Area of waterplane(Awl) should not be less than a limit about 120 m2 in


order to avoid enormous changes to draft as mentioned above.

 For practical reasons Bstr should be less than 1.6 m (and for easier
success of the constrain above)

 Draft=5 m (constant)

 Lstr_n,Lstr_t : should be a rate of Lgn,Lgt (no more than 30-60 %,


depending on radious) for practical reasons ( to have a clear flow at
propeller etc)

Something that must be stated here is that the control of geometric


irregularities occur very often in the case of arbitrary parametric design where
large variation is required. So these constraints ensure that the developed
geometry will remain within feasible boundaries regarding its shape and
contributes in avoiding system crashes which are not rare when dealing with
sophisticated holistic systems.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

64
5.5.1 DESIGN OF EXPERIMENT (DOE)

As referred and above the optimization initiated with a design engine called
“exhaustive search”. This engine just asks two things. The first one is the
range of the parameters and the second the subdivisions. The subdivision is a
number that defines how many equal spacings the range of the parameter will
have. So this is the number of models that are created.

After that the “Sobol” design engine was called. Firstly with “Sobol” a
verification of the results of “exhaustive search” was performed. After that a
deeper investigation of the design space was made. So the tendency of every
single parameter was understood in detail.

We need to mention that R15,5Kn is the total resistance of the vessel at


15,5Knots and R31Kn is the total resistance of the vessel at 31Knots. Both
are measured in KiloNewtons(KN).

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

65
Here is presented the trend of every parameter for the low and the high
speed:

 Radmid= Displacement distribution parameter(for parameter value


=1 there is a parallel midbody, for parameter value>1 there is a
displacement distribution to the centrer,along the length, and for
parameter value <1 there is a displacement distribution to the
edge ,along the length).

R15,5Kn-Radmid
200
190
180
170
160
Rt(KN)

150
140
130
120
110
100
0,5 0,6 0,7 0,8 0,9 1 1,1 1,2 1,3 1,4
Radmid value(non dimentional)

Graph 1 Rt-Radmid

R31Kn-Radmid
660
655
650
645
Rt(KN)

640
635
630
625
620
0,5 0,6 0,7 0,8 0,9 1 1,1 1,2 1,3 1,4
Radmid value(non dimentional)

Graph 2 Rt-Radmid

It is obvious that for the low speed an increase at radmid decreases the total
resistance. For the high speed the converse it is happening.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

66
 Bstrut= Strut beam

Bstr-R15,5Kn
155

150

145
Rt(KN)

140

135

130

125
1 1,2 1,4 1,6 1,8 2 2,2 2,4 2,6
Bstr Value(m)

Graph 3 Rt-Bstr

Bstr-R31Kn
680
670
660
650
Rt(KN)

640
630
620
610
1 1,2 1,4 1,6 1,8 2 2,2 2,4 2,6
Bstr Value(m)

Graph 4 Rt-Bstr

Here it is clear for both speeds that an increase to Bstr increases the total
resistance.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

67
 Latsep= half distance between lower hulls

R15,5Kn-latsep
200
190
180
170
160
Rt(KN)

150
140
130
120
110
100
2 3 4 5 6 7 8 9
Latsep value(m)

Graph 5 Rt-Latsep

R31Kn-Latsep
750

700

650
Rt(KN)

600

550

500
2 3 4 5 6 7 8 9
Latsep value(m)

Graph 6 Rt-Latsep

About Latsep we can notice the following:For the low speed and within the
given logical range we see a total minimum that is about at 5m. So the
optimum distance between lower hulls is about 10m .
For the high speed there is no minimum but the total resistance decreases
with an increase at Latsep. The point is that we cannot push this parameter a
lot because structural problems will turn up.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

68
 Lengthgondola= lower hull length

R15,5Kn-Lgon
220

200

180
Rt(KN)

160

140

120

100
40 42 44 46 48 50 52 54 56
Lgon value (m)

Graph 7 Rt-Lgon

R31Kn-Lgon
740
720
700
Rt(KN)

680
660
640
620
600
40 42 44 46 48 50 52 54 56
Lgon value (m)

Graph 8 Rt-Lgon

Here we observe for the low speed that there are some local minimum and a
total minimum for least total resistance that is about at parameter value 46 m.
Now for the high speed beyond some irregularities at 44 m an increase at
Lengthgondola (essentially at its parallel body) decreases the total resistance.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

69
 Lgn=lower hull nose length

R15,5Kn-Lgn
160
155
150
145
Rt(KN)

140
135
130
125
3 4 5 6 7 8 9 10 11
Lgn value (m)

Graph 9 Rt-Lgn

R31Kn-Lgn
654
652
650
648
646
Rt(KN)

644
642
640
638
636
634
632
3 4 5 6 7 8 9 10 11
Lgn value (m)

Graph 10 Rt-Lgn

About Lgn we can notice that generally in low speed as we increase the
parameter value total resistance decreases. On the contrary at the high speed
the opposite is happening so an increase at Lgn increases the total
resistance.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

70
 Lgt= lower hull tail length

R15,5Kn-Lgt
190
180
170
160
Rt(KN)

150
140
130
120
110
100
4 5 6 7 8 9 10 11 12
Lgt value (m)

Graph 11 Rt-Lgt

R31Kn-Lgt
644
642
640
638
Rt(KN)

636
634
632
630
628
626
4 5 6 7 8 9 10 11 12
Lgt value(m)

Graph 12 Rt-Lgt

Here as for Lgt the results are almost the same with the previous one
(Lgn).Apart from some little irregularities, at low speed an increase to
the Lgt decreases the total resistance. On the contrary at the high
speed the opposite is happening again.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

71
 Lstr_n=strut nose length

R15,5Kn-Lstr_n
142,5
142
141,5
141
140,5
Rt(KN)

140
139,5
139
138,5
138
137,5
137
1 2 3 4 5 6 7
Lstr_n value (m)

Graph 13 Rt-Lstr_n

R31Kn-Lstr_n
644
643
642
Rt(KN)

641
640
639
638
637
1 2 3 4 5 6 7
Lstr_n value (m)

Graph 14 Rt-Lstr_n

Lstr_n seems to decrease for minimum total resistance at low speed


and to increase at high speed. But this tendency is not so obvious.
Generally as we can see a great change at its value has very little
response to the total resistance.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

72
 Lstr_t=strut tail length

R15,5Kn-Lstr_t
144
142
140
Rt(KN)

138
136
134
132
1 2 3 4 5 6 7 8
Lstr_t value (m)

Graph 15 Rt-Lstr_t

R31Kn-Lstr_t
644

642

640
Rt(KN)

638

636

634
1 2 3 4 5 6 7 8
Lstr_t value (m)

Graph 16 Rt-Lstr_t

Lastly Lstr_t tends to decrease for least total resistance for low speed.
About the high speed a total maximum seems to be created at from 3 to 5 m
so this area should be avoided. But again as Lstr_n the response to the total
resistance is least . That is the reason which other factors may overpass this
restricted area. As mentioned earlier we need a clean flow at the propeller so
Lstr_t must be a rate of Lgt (about 30-60%)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

73
5.5.2 MULTI OBJECTIVE OPTIMIZATION-GENETIC ALGORITHMS

Since the design of experiment has provided us with a lot of feedback


regarding the response of the Swath while varying, it is time now to proceed
to the formal multi-objective optimization. The range of the parameters has
been reduced after a thorough examination of the best designs of the DoE. As
a result, a region of good variants within the space of feasible design has
been identified. In this area, the optimization will be executed in order to find
the best design and provide the Pareto-Frontier. The NSGA-II will be utilized
to perform the multi-criteria optimization. The genetic algorithm will run many
times in order to within a reduced better range each time. That will not only
find more optimal designs but it will also make sure that the genetic algorithm
will not be trapped to local minimum( often due to the elitism).
The same procedure will be followed with the MOSA algorithm to be
absolutely sure about the results .

Generally the boundaries of the parameters was given to the algorithm as


follows:

Table 1 Parameter boundaries for optimization

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

74
Evaluation of the results. (NSGA II).

The range of the parameters and the objectives are listed below:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

75
Table 2 Boundaries of parameter and evaluations (NSGA-II)

The following diagram depict the scatter of the objectives.It is shown that the
best compromises have been achieved since the Pareto-Frontier is apparent.

240

220

200

180
R15,5(KN)

160

140

120

100
500 520 540 560 580 600 620 640 660 680 700
R31(KN)

All Designs BestR15.5(mostly)


Best equal BestR31(mostly)
BestR31 BestR15.5

Graph 17: Evaluations R15,5 vs R31 (NSGA-II)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

76
It should be mentioned here that the non-feasible designs were excluded from
the above diagram. In every diagram, the reference design is not close to the
Pareto-Frontier in terms of the objectives. That proves the fact that there was
a merit, and that the genetic algorithm has completed its task successfully.
Therefore it can be assumed that the NSGA II found the best compromises for
the two objectives.

Evaluation of the results. (MOSA)

The range of the parameters and the objectives are listed below:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

77
Table 3 Boundaries of parameter and evaluations (MOSA)

The following diagram depict the scatter of the objectives .It is shown that the
best compromises have been achieved since the Pareto-Frontier is apparent.

MOSA R15,5 Kn vs R31 Kn

240

220

200
R15,5(KN)

180

160

140

120

100
500 520 540 560 580 600 620 640
R31(KN)

All Designs BestR15.5(mostly)


Best equal BestR31(mostly)
BestR31 BestR15.5

Graph 18: Evaluations R15,5 vs R31 (MOSA)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

78
It should be mentioned here that the non-feasible designs were excluded from
the above diagram. In every diagram, the reference design is not close to the
Pareto-Frontier in terms of the objectives. That proves the fact that there was
a merit, and that the genetic algorithm has completed its task successfully.
Therefore it can be assumed that the MOSA found the best compromises for
the two objectives.

For reasons of completeness some diagrams from previous runs are


presented here.

1,05
1
0,95
R15,5(non-dimensional)

0,9
0,85
0,8
0,75
0,7
0,65
0,6
0,55
0,8 0,85 0,9 0,95 1 1,05
R31(non-dimensional)

Graph 19: Evaluations R15,5 vs R31

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

79
1

R15,5(non-dimensional) 0,9

0,8

0,7

0,6

0,5

0,4
0,8 0,85 0,9 0,95 1
R31(non-dimensional)

Graph 20: Evaluations R15,5 vs R31

240

220

200
R15,5(KN)

180

160

140

120

100
500 550 600 650 700
R31(KN)

Graph 21: Evaluations R15,5 vs R31

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

80
5.6 Best Designs

The multi-objective optimization created a wide variety of feasible designs. It


also achieved to develop designs of best compromise, while keeping the
constraints in an acceptable range. Pareto frontiers in the above diagrams
prove this fact. In order to choose some the best designs, an objective
function must be introduced in order to meet specific needs. There are formal
procedures for this task like Multiple Attribute Decision Making (MADM)
presented by Sen and Yang. In this case, a manual exploitation of the design
space with the objective functions is conducted.

Regarding with the objective function we chose three best designs that have
already been shown in the above diagrams (NSGA II).

The first one is based 80% at the low speed and 20% at the high speed.
The second one is based 50% at the low speed and 50% at the high speed.
These two coincides to be the same design and is as follows:

Table 4 Parameter and evaluation values of mostly R155 and best equal (NSGA-II)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

81
Figure 5.10 Best design for low (and equal) speed (NSGA-II)

The third one is based 20% at the low speed and 80% at the high speed and
is as follows:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

82
Table 5 Parameter and evaluation values of mostly R31 (NSGA-II)

Figure 5.11 Best design for high speed(NSGA-II)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

83
Regarding with the objective function we chose three best designs that have
already been shown in the above diagrams (MOSA).

The first one is based again 80% at the low speed and 20% at the high speed
and is as follows:

Table 6 Parameter and evaluation values of mostly R155(MOSA)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

84
Figure 5.12 Best design for low speed(MOSA)

The second one is based again 20% at the low speed and 80% at the high
speed and is as follows:

Table 7 Parameter and evaluation values of mostly R31(MOSA)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

85
Figure 5.13 Best design for high speed(MOSA)

The third one is based again 50% at the low speed and 50% at the high
speed and is as follows:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

86
Table 8 Parameter and evaluation values of best equal (MOSA)

Figure 5.13 Best design for moderate speed(MOSA)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

87
5.7 VALIDATING RESULTS WITH SHIPFLOW

Taking into account that the MICHLET is a very fast code that calculates the
total resistance within seconds for a range of speeds , a validation of the
results was needed to be fulfilled. So SHIPFLOW had to be introduced. There
is no need to run again optimization with SHIPFLOW as this would take a lot
of time and computational power. That was the reason that was decided to
extract the best designs from FRIENDSHIP FRAMEWORK, alternate to a
form that is understood by SHIPFLOW and run them again to check the
results. At this point we should state that this was not an easy job as the best
designs refer to a particular speed. So the CFD code had difficulties in giving
objective results for the range of speed needed. For example the best design
for the low speed had very good results for speed of 15,5Kn but also gave us
poor results for the speed of 31Kn.

The only solution was to change the gridlines at SHIPFLOW configuration.


With some iterations correcting gridlines the results were far more better but
again not absolutely cogent.

Here are presented the results of the SHIPFLOW for each of the five best
designs from MOSA(the three presented above, the best for the low speed
and the best for the high speed). At each of the best designs some
intermediate speeds were also calculated in order to create a typical
resistance curve (at least for a small range of speeds).

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

88
 Best R155

Rt-V(Kn)Shipflow
Rt-V Michlet-R15,5
MichletR31
700
600
500
400
Rt(KN)

300
200
100
0
14 19 24 29
V(Kn)

Graph 19: Rt –V (Shipflow)

For the best design at the speed of 15,5 Kn we can notice that there is a
convergence to the codes at this speed but as it raises divergence is
inevitable. That is why at the high speed SHIPFLOW did not converged to
a Cw it created high waves and the iterations stopped by the program
itself.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

89
 BestR155(mostly)

Rt-V(Kn)Shipflow
Rt-V Michlet-R15,5
MichletR31
600

500

400
Rt(KN)

300

200

100

0
14 19 24 29
V(Kn)

Graph 20: Rt –V (Shipflow)

Here we can see exactly what was mentioned to the previous design but
obviously to a greater scale. Again convergence at the low speed and
divergence at the high one. The reasons still the same.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

90
 Best equal

Rt-V(Kn)Shipflow
Rt-V Michlet-R15,5
MichletR31
600

500

400
Rt(KN)

300

200

100

0
14 19 24 29
V(Kn)

Graph 21: Rt –V (Shipflow)

At the best design for both speeds the above phenomenon is starting to
moderate. For example at the high speed SHIPFLOW ran more iterations for
the previous designs. Again we cannot extract safe results but at least we
have something to base on.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

91
 BestR31(mostly)

Rt-V(Kn)Shipflow
Rt-V Michlet-R15,5
MichletR31
600

500

400
Rt(KN)

300

200

100

0
14 19 24 29
V(Kn)
Graph 22: Rt –V (Shipflow)

At this design we witness another phenomenon of the previous. SHIPFLOW


stopped running because it finished the iterations but it did not achieved
convergence. This happened at both speeds.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

92
 Best R31

Rt-V(Kn)Shipflow
Rt –V Michlet-R15,5
MichletR31
600

500

400
Rt(KN)

300

200

100

0
14 19 24 29
V(Kn)

Graph 23: Rt –V (Shipflow)

At the best design for the high speed we can observe that SHIPFLOW
diverged at slow speed(stopped by iterations) but converged at the high
speed( the reverse phenomenon for the first two designs)

Generally we can say the SHIPFLOW always underestimated the total


resistance by the aforementioned ways. A number of 25 runs were made to
SHIPFLOW( five designs for five speeds each). The runs that did not
converged( they had their iterations stopped by the program) had their
gridlines changed and ran again. This procedure stopped when almost every
run was successful ( at least they stopped by the iterations , not by the
program).

The last but not least that has to be mentioned is that indeed there is a
deviation regarding to the two programs. Happily this deviation is smaller
where the SHIPFLOW converged and greater where it diverged. Something
that makes perfect sense. Unfortunately as we already said SHIPFLOW
underestimate the total resistance( hence the wave resistance, as the friction
is calculated by ITTC and is the same for both programs). This cannot be
explained at the boundaries of this diploma thesis but the answer may lies at
the code of each program itself or at the nature of the problem proposed
(Swath is a special kind of ship)

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

93
5.8 SINGLE OBJECTIVE OPTIMIZATION

At this point a parenthesis should be opened. As mentioned above at the


multi-objective optimization the parameters are usually contradictory. With
the optimization problem as stated above two single objective optimizations
were accomplished in order to explore how are our parameters vary when our
goal lies only at the slow or only at the fast speed. In this case we get
released from the contradictory parameters, and as a result we can get even
better results for every single speed. Something that has to be reffered at this
point is that the parameters has been changed regarding with the DOE for
each speed in order to find the most optimized design.

The results for the low speed are as follows:

Table 9 Boundaries of parameters and evaluations

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

94
It is obvious that if we guide the parameters at their optimal values the results
we take are far more than satisfying. Here we can see that the total resistance
for the best design for the slow speed can be reduced even more and catch a
value up to 88KN. So if we decide that the we finally run at the speed of
15,5Kn we then have ready our optimized Swath

Now if we change our minds and switch at the high speed the results are
again very promising, as the total resistance can be reduced at 503KN:

Table 10 Boundaries of parameters and evaluations

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

95
5.9 Comparison of AQ3 with AQ

Since this diploma thesis started with the data of the Passenger Ferry Aegean
Queen (AQ, [13]), a comparison of the herein obtained resistance results
should be made with those of AQ. So from the resistance curve of AQ the
total resistance was found for two significant speeds. All results are compared
at the following table:

AQ Best Mostly Best Mostly Best


Rt(KN) AQ (tests) (predictions) R155 R155 equal R31 R31
For the speed
of 15.5 Kn 250 260 113 119 138 169 212
For the speed
of 31 Kn 645 565 642 568 541 524 516

Table 11 Compared results of AQ & AQ3

Table 11 compares the total resistance data of AQ( test results and numerical
predictions) with the total resistance data of AQ3 ( the best five designs as
resulted from the optimization and presented at an earlier part of this diploma
thesis) . As explained before:
Best R155 is the best design for the low speed ,
Mostly R155 is the best design for mostly the low speed,
Best equal is the best design for both speeds,
Mostly R31 is the best design for mostly the high speed,
Best R31 is the best design for the high speed.
From table 11 it is obvious that the models proposed here are much more
efficient than AQ.
In this comparison however some important factors should be mentioned:

 The predicted values of AQ at the high speed proved much smaller


compared to the experimental values, whereas the comparison at low
speed is much better. This is attributed to various non-potential flow
phenomena, like flow separation, but also the change of the mean draft
and trim of the vessel that was not accounted for in the theoretical
predictions..

 AQ had a constrained lower hull(gondola) and strut to fit the engines.


This increases the total resistance of AQ compared to an
unconstrained optimized design. AQ3 instead does not require this
adjustment due to the fact that by the technological improvements of
diesel engines of same power, they now fit in much lesser space (see
characteristics of MTU engines, e.g. 20V 1163 TB73 ).

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

96
6. Conclusions

The work presented herein, demonstrated the applicability of an advanced


technology ship design approach using parametric design tools for
optimization at the conceptual design phase.

The preliminary design phase of Swath has been realized, utilizing parametric
modeling tools in the FRAMEWORK of simulation-driven design.

In this case study, it has been achieved to build up a robust optimization


model. Several sub-systems have been developed in order to cover many
aspects of the design optimization problem that compose a fully automated
package regarding the design of a novel, unconventional catamaran.

The core of this method is found in the fully parametric model, which is
applicable to a wide range of global dimensions and local characteristics,
retaining its fairness of shape and feasibility of its properties. It is generated
using the tight coupling of the computer aided engineering tool FRIENDSHIP-
FRAMEWORK, and the computer workbench MICHLET, (and the flow solver
SHIPLFLOW).

Since the designer has developed all the subsystems and has carefully
examined the interaction and dependencies that occur between the different
factors, an extensive parametric variation study was undertaken in order to
explore the feasible boundaries of a multi-dimensional design space. The final
stage of multi-objective optimization, led by two Genetic Algorithms provided
many favourable designs with rather competitive characteristics compared to
existing ships of the same type and range, and other conceptual designs.

Regarding the investigated design concept, it was shown, that a special type
of ship , a Swath, that is fully unconventional and without a parallel midbody
has advantageous characteristics regarding powering demands from all
aspects as shown above.

The greater advantage of all this coupling is time. There are so many design-
optimization packages most of them trying to achieve as much precision as
possible. The concept of this thesis is not precision but time. After the design
process it can give the first results within hours without any special computer
powering.

So the contribution of this thesis, while examining the overall work is as


follows:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

97
 Development and application of a robust and very fast global
optimization method for Swath design.
 Adoption of the method within one program (FRIENDSHIP
FRAMEWORK) that is used for the geometrical modeling, simulation
and optimization, using only MICHLET as an external software.
 Analysis of design variables sensitivities and use of them as design
directives for a quick dimensioning during the preliminary design.
 Initial, global optimization of a Swath using the MOSA and NSGAII
algorithm.
 Promising results of Swath design

Still, there are many unexplored regions. In order to achieve a greater degree
of holism and improve the decision making process at the preliminary phase,
more aspects of the ship design problem have to be integrated in the
automated optimization process.

It seems that the multi-discipline task of ship design enters a new era, where
the naval architect and the designer will have to embrace a totally new
perspective. Parametric modeling and simulation-driven design have the
potential to change radically the traditional way of thinking and acting in
marine industry.

Finally some future perspectives are presented below for further work:

 Taking into consideration all these more work can be done when
focusing more at the design process including all hydrostatic
characteristics ,stability and damage stability and so on.
 Designing and optimizing the structural part of the vessel for least
weight , better moments of inertia etc.
 Exploring the seakeeping characteristics especially at high speeds and
rough seas, utilizing stabilizing fins.
 Economic and technical study with statistical data and risks
assessments, adjusting the inner space in order to maximize the
capacity of passengers and cars.
 Searching for a special type of propeller with better efficiency, studying
simultaneously the wake around it.
 Converting the Swath to Slice and comparing the two types.
 Getting away from the conventional catamaran- Swath design and
keeping this special hull form for more hulls so that wave resistance
for some of the hulls can be greatly diminish.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

98
7. REFERENCES

1. Bertram V. “Practical Ship Hydrodynamics.” Butterworth-Heinemann, Oxford, 2000.

2. Boulougouris, E., Papanikolaou, A., “Preliminary Design of a SWATH Corvette for Operations in the
Littoral Environment”, Proc. Int. Symposium WARSHIP 2001 on Future Surface Warships, RINA,
London, June 2001.

3. Chotzopoulos D. “Πολςκπηηιπιακη βεληιζηοποιηζη μελεηηρ δεξαμενοπλοιυν με οικονομοηεσνικα


κπιηηπια και σπηζη γενεηικυν αλγοπιθμυν” ,National Technical University of Athens , 2011

4. Deb, K.: “A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II, IEEE transactions on
evolutionary computation”, vol. 6, no.2, April 2002

5. Tuck E. O. and Lazauskas L. “Optimum hull spacing of a family of multihulls.” University of Adelaide ,
1998

6. Tuck E.O. “Wave Resistance of Thin Ships and Catamarans” University of Adelaide , 1997

7. FRIENDSHIP-SYSTEMS, Potsdam. FRIENDSHIP-FRAMEWORK User Guide, Nov. 2009

8. FRIENDSHIP-SYSTEMS, Potsdam. FRIENDSHIP-FRAMEWORK, issue tracker

9. Herbert J.Koelman: “Computer Support for Design, Engineering and Prototyping of the Shape
of Ship Hulls.” 1999

10. http://thesis.library.caltech.edu/1151/1/Wilson_mb_1971.pdf

11. http://www.boatdesign.net/forums

12. http://www.cyberiad.net/leo.htm

13. http://www.friendship-systems.com/forum/

14. http://www.naval-architecture.co.uk

15. Koutroukis G. “Parametric design and multi-objective optimization-study of an ellipsoidal containership” ,


National Technical University of Athens , 2012

16. Lazauskas L. , Michlet user manual, 2010

17. Nikolopoulos L. “A holistic methodology for the optimization of tanker design and operation and
its applications.” , National Technical University of Athens , 2012

18. Pallas G. “Απιθμηηικη και πειπαμαηικη διεπεςνηζη ανηιζηαζηρ και ςτοςρ κςμαηορ διγαζηπος
ζκαθοςρ συπιρ βολβo.” , National Technical University of Athens , 2013

19. Papanikolaou A. , “Μελέηη Πλοίος, Μεθοδολογίερ Ππομελέηηρ”, Σεύσορ 1,Εκδ. ΢ςμεών, 2009

20. Papanikolaou, Α. , Zaraphonitis, G., Androulakakis, Μ., "Hydrodynamic Aspects and Conceptual
Design of SWATH vessels", Final Report to the Greek Secr. General for Res. and Tech.,ΠΡΟΠΕ 86Γ924,
NTUA, March 1990 (Greece).

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

99
21. Papanikolaou, Α. D., Bouliaris, Ν., Koskinas, C., Pigounakis, Κ., "Conceptual Design and
Hydrodynamic Analysis of SMUCC - SWATH Multipurpose Container Carrier for Short Sea
Shipping", Proc. 3rd Int. Conf. on Fast Sea Transportation, FAST' 95, Sept.1995, Travemuende
(Germany).

22. Papanikolaou, Α., "Computer-Aided Preliminary Design of a High-Speed SWATH Passenger/Car


Ferry", Proc. of 4th IMSDC '91 Conf., Kobe, May 1991 (Japan).

23. Papanikolaou, Α., "EU409 EUROMAR Project : SMURV-SWATH Multipurpose Research Vessel
for the Mediterranean Area", NTUA-EU409 3rd Report, Athens, February 1992, (Greece).

24. Papanikolaou, Α., "Hydrodynamic Aspects and Conceptual Design of a SWATH-Passenger/Car


Ferry", Journal Technica Italiana, Vol. 52,1988 (Italy).

25. Papanikolaou, Α., Androulakakis, Μ., "Hydrodynamic Optimization of High-Speed SWATH",


Proc. of 1st FAST '91 Conf., Trondheim, June 1991, (Norway).

26. Papanikolaou, Α., Atlar, Μ., Khattab, Ο., "Hydrodynamic Analysis and Design of a SWATH
Multipurpose Research Vessel", Proc. 2nd Int. Conf. on Fast Sea Transportation FAST'93,
Yokohama, December 1993 (Japan).

27. Papanikolaou, Α., Nowacki, Η. et al., "Concept Design and Optimization of α SWATH
Passenger/Car Ferry", Proc. IMAS-89 Int. Conf. on Applications of New Technology in Shipping,
Athens, May 1989, (United Kingdom).

28. Papanikolaou, Α., Zaraphonitis, G., Androulakakis, Μ., "Preliminary Design of a High-Speed
SWATH Passenger Car/ Ferry", Journal Marine Technology, SNAME, Vol. 28, Νο. 3, ππ.129-141,
May 1991 (USA).

29. Papanikolaou, Α., Zaraphonitis, G., Koskinas, C., Sawas, Ι., "Stability of a SWATH car/ ferry in
calm water and in waves", Proc. STAB '90 Conf., Naples, Sept.1990 (Italy).

30. Rizos H. , “Mελεηη ε/γ- ο/γ SWATH για ηην ελληνικη ακηοπλοια” , National Technical University
of Athens , 2005

31. Skoupas S. “Αναπηςξη μεθοδολογιαρ για ηην μελεηη και ζσεδιαζη – βεληιζηοποιηζη επιβαηηγυν –
οσημαηαγυγυν πλοιυν νεαρ ηεσνολογιαρ.” , National Technical University of Athens , 2011

32. Wikipedia http://en.wikipedia.org/wiki/Wiki

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

100
APPENDICES

APPENDIX A - TriSwath

The aim of this annex is to optimize the placement of individual hulls in a


multihull configuration, in order to minimize wavemaking. No attempt is made
here to optimize the shape of individual hulls. For a general multihull vessel,
these are coupled tasks, but we consider here only a special case where they
are uncoupled.

The name of the three-hull designed is TriSwath and the geometry is the
same as the Swath designed at this diploma thesis but scaled by a factor of
0.66 in order to keep the displacement fixed at 1000t. The formation of the
vessel is like an arrow and the placement of the individual hulls is as follows:

Figure A.1 TriSwath hull placement [19]

The above geometry was designed at FRIENDSHIP-FRAMEWORK and a


DOE ran. As stated before the form of the hull did not touched but only the
placement. That left only two parameters to explore. The lateral and the
longitudinal separation. So Sobol algorithm was utilized with these parameters
and evaluations the total resistance at the same speeds as the AQ3.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

101
Attribute Active Active
Name lat_sep long_sep eval_R155_tri eval_R31_tri
Scope |parameters |parameters
Reference
Lower Bound 3.15625 10.3125 155.25453 593.13305
Upper Bound 12.84375 29.6875 480.06862 947.44427

Feasible Designs: 100 %


Mean Utilization Index
Mean 7.9727746 19.983428 266.73609 829.11038
Sample Standard Deviation 2.8790697 5.7531561 72.029338 65.041477
Error-free: 100 % 100% 100% 100% 100%
Table A.1 Boundaries of parameters and evaluations

As we can see it is close to the best optimized values from AQ3. And this
without a form optimization. Moreover it must not be forgotten that obviously
the wetted surface is more than AQ3 something that means increased friction
resistance. But the most interesting and promising at the same time is that the
speeds we assumed is a random(the same as AQ3). At these exact speed the
resistance curve does not present a local minimum(unlikely it is near a local
maximum). So if a it is about to create a new model and we have the right to
adjust the speed we need to run the results are awesome.
The resistance curve is as follows (from a random model of the Sobol results):

Graph A.1

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

102
Finally the trends of the longitudinal and lateral separation are presented:

Graph A.2

The trend of the lateral separation here is not so obvious although as the
parameter increases the Rt for 15,5Kn is stabilizing.

Graph A.3
It is obvious at the diagram that the least Rt for 15,5Kn is for parameter value
of 17-21m

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

103
Graph A.4

The lateral separation for the high speed seems to be almost the same with
the low speed. Just the scatter extends to a lesser degree. Agan the trend is
not so obvious.

Graph A.5
Not the same with the low speed , although Rt at 31 Kn seems to decrease at
the greater values of the longitudinal separation.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

104
APPENDIX B - DiamondSwath

The aim of this annex is again to optimize the placement of individual hulls in
a multihull configuration, in order to minimize wavemaking. No attempt is
made here to optimize the shape of individual hulls. For a general multihull
vessel, these are coupled tasks, but we consider here only a special case
where they are uncoupled.

The name of the tetra-hull designed is DiamondSwath and the geometry is the
same as the Swath designed at this diploma thesis but scaled by a factor of
0.5 in order to keep the displacement fixed at 1000t. The formation of the
vessel is like an diamond configuration while all hulls are identical. The
placement of the individual hulls is as follows:

Figure B.1 DiamondSwath hull placement [19]

The above geometry was designed at Friendship-FRAMEWORK and a DOE


ran. As stated before the form of the hull did not touched but only the
placement. That left again only three parameters to explore. The lateral
separation, the longitudinal separation and the longitudinal separation of the
last hulls behind the others. In order for the optimization to be least complex
the third parameters was considered as the longitudinal separation multiplied
by two. So Sobol algorithm was utilized with these two parameters.
Evaluations were again the total resistance at the almost the same speeds as
the AQ3 slightly changed the slow one.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

105
Attribute Active Active
Name lat_sep long_sep eval_R18 eval_R31
Scope |parameters |parameters
Reference
Lower Bound 3.1875 10.78125 136.20971 525.27443
Upper Bound 14.4375 34.609375 808.48636 1012.3346

Feasible Designs: 100 %


Mean Utilization Index
Mean 9.0392442 23.108648 428.74786 888.01159
Sample Standard Deviation 3.2231614 7.0096717 236.74881 113.50315
Error-free: 100 % 100% 100% 100% 100%
Table B.1 Boundaries of parameters and evaluations

The results here seems to be more promising as the total resistance was
reduced even more (without a form optimization). And it is easily understood
that the friction resistance is far more greater. But now indeed we are to local
minimum of the resistance curve (as stated below). Anyway the evaluations
can also be diminished more.

Graph B.1

Finally the trends of the two parameters are stated below:

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

106
Graph B.2

Lateral separation seems to decrease for diminished total resistance at 18Kn

Graph B.3

Longitudinal separation tends to be around 25 m for least total resistance at


18 Kn.

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

107
Graph B.4

The trend here is not so obvious but some good designs and their values give
a hint.

Again without being obvious the least total resistance for the speed of 31 Kn
seems to be at very low or very high longitudinal separation values.
Graph B.5

C. Papandreou, Parametric Design and Multi-objective Optimization of SWATH


Diploma Thesis, NTUA, Ship Design Laboratory, June 2013

108

You might also like