Polish Maritime Research

Address of Publisher
& Editor's Ofce :

GDASK UNIVERSITY
OF TECHNOLOGY
Faculty
of Ocean Engineering
& Ship Technology
ul. Narutowicza 11/12
80-952 Gdask, POLAND
tel.: +48 58 347 17 93
fax : +48 58 341 47 12
e-mail : [email protected]
Account number :
BANK ZACHODNI WBK S.A.
I Oddzia w Gdasku
41 1090 1098 0000 0000 0901 5569
Editorial Staff :
Witold Kirkor Editor in Chief
Przemysaw Wierzchowski Scientic Editor
Maciej Pawowski Editor for review matters
Tadeusz Borzcki Editor for international relations
Cezary Spigarski Computer Design
Domestic price :
single issue : 20 z
Prices for abroad :
single issue :
- in Europe EURO 15
- overseas US$ 20
ISSN 1233-2585
NAVAL ARCHITECTURE
3 CZESAW DYMARSKI, RAFA ROLBIECKI
Comparative analysis of selected
design variants of propulsion system
for an inland waterways ship
8 PAWE DYMARSKI
Numerical simulation
of viscid ow around hydrofoil
MARINE ENGINEERING
16 JUSTYNA LZAK-ONA
On the application of the articial neural
network method to a neural simulator
of steam turbine power plant
OPERATION & ECONOMY
22 TOMASZ CEPOWSKI
Approximation of pitching motion of S-175
containership in irregular waves
on the basis of ships service parameters
26 MAREK NARKIEWICZ
Assessment of the safety level
of a ships passing manoeuvre in the fairway
POLISH
MARITIME
RESEARCH
in internet
www.bg.pg.gda.pl/pmr.html
Index and abstracts
of the papers
1994 2005
PUBLISHER :
CONTENTS
POLISH MARITIME RESEARCH
No 1(47) 2006 Vol 13
The papers published in this issue have been reviewed by :
Prof. J. Girtler ; Prof. S. Gucma
Assoc.Prof. K. Kosowski ; Prof. J. Szantyr
Prof. T. Szelangiewicz
POLISH MARITIME RESEARCH is a scientic journal of worldwide circulation. The journal appears
as a quarterly four times a year. The rst issue of it was published in September 1994. Its main aim is to
present original, innovative scientic ideas and Research & Development achievements in the eld of :
Engineering, Computing & Technology, Mechanical Engineering,
which could nd applications in the broad domain of maritime economy. Hence there are published papers
which concern methods of the designing, manufacturing and operating processes of such technical objects
and devices as : ships, port equipment, ocean engineering units, underwater vehicles and equipment as
well as harbour facilities, with accounting for marine environment protection.
The Editors of POLISH MARITIME RESEARCH make also efforts to present problems dealing with
education of engineers and scientic and teaching personnel. As a rule, the basic papers are supplemented
by information on conferences , important scientic events as well as cooperation in carrying out interna-
tional scientic research projects.
Editorial
Scientic Board
Chairman : Prof. JERZY GIRTLER - Gdask University of Technology, Poland
Vice-chairman : Prof. ANTONI JANKOWSKI - Institute of Aeronautics, Poland
Vice-chairman : Prof. MIROSAW L. WYSZYSKI - University of Birmingham, United Kingdom
Prof. ANTONI ISKRA
Pozna University of Technology
Poland
Prof. YASUHIKO OHTA
Nagoya Institute of Technology
Japan
Dr POUL ANDERSEN
Technical University of Denmark
Denmark
Prof. JAN KICISKI
Institute of Fluid-Flow Machinery
of PASci
Poland
Prof. ANTONI K. OPPENHEIM
University of California
Berkeley, CA
USA
Dr MEHMET ATLAR
University
of Newcastle
United Kingdom
Prof. KRZYSZTOF ROSOCHOWICZ
Gdask University
of Technology
Poland
Prof. ZYGMUNT KITOWSKI
Naval University
Poland
Prof. GRAN BARK
Chalmers University
of Technology
Sweden
Prof. KLAUS SCHIER
University of Applied Sciences
Germany
Prof. WACAW KOLLEK
Wrocaw University of Technology
Poland
Prof. MUSTAFA BAYHAN
Sleyman Demirel University
Turkey
Prof. ODD M. FALTINSEN
Norwegian University
of Science and Technology
Norway
Prof. FREDERICK STERN
University of Iowa,
IA, USA
Prof. NICOS LADOMMATOS
University College
London
United Kingdom
Prof. PATRICK V. FARRELL
University of Wisconsin
Madison, WI
USA
Prof. JZEF SZALA
Bydgoszcz University
of Technology and Agriculture
Poland
Prof. JZEF LISOWSKI
Gdynia Maritime
University
Poland
Prof. STANISAW GUCMA
Maritime University
of Szczecin
Poland
Prof. JERZY MATUSIAK
Helsinki University
of Technology
Finland
Prof. JAN SZANTYR
Gdask University
of Technology
Poland
Prof. MIECZYSAW HANN
Technical University of Szczecin
Poland
Prof. EUGEN NEGRUS
University of Bucharest
Romania
Prof. BORIS A. TIKHOMIROV
State Marine University
of St. Petersburg
Russia
Prof. DRACOS VASSALOS
University of Glasgow and Strathclyde
United Kingdom
Prof. KRZYSZTOF WIERZCHOLSKI
Gdask University of Technology
Poland
3
POLISH MARITIME RESEARCH, No 1/2006
INTRODUCTION
Today several kinds of propulsion systems are applied on
board inland waterways ships, out of which the following can
be enumerated :
+ conventional one tted with a combustion engine, toothed
gear and xed or controllable pitch propeller, free or ducted
in a xed or pivotable Korts nozzle
+ combustion-electric one tted with an electric transmission
and frequency converter making it possible to steplessly
control rotational speed of the propeller which may be
xed one
+ combustion-hydraulic one tted with a hydrostatic trans-
mission and xed propeller
+ propulsion system tted with two azimuthal propellers
(rotatable thrusters) driven by combustion engines through
a toothed, electric or hydrostatic transmission
+ propulsion system tted with cycloidal (Voith- Schneider)
propellers
+ propulsion system tted with water jet propellers.
Construction of inland waterways ships and their propulsion
systems decisively depend on depth of waterways in which
a given ship is to sail, as well as on dimensions of sluices exi-
sting in the waterways, and also on other conditions including
ecological ones.
Below are presented basic design assumptions and selected
design concepts of propulsion system for an inland waterways
passenger ship intended for sailing on a shipping route of the
minimum water depth of 1.2 m.
Comparative analysis of selected design variants
of propulsion system for an inland waterways ship
Czesaw Dymarski
Rafa Rolbiecki
ABSTRACT
In this paper are presented design assumptions and technical conditions as well as selected design versions
of propulsion system for an inland waterways ship, and also a preliminary comparative analysis of two
solutions. In the rst version this is a combustion-electric systemtted with frequency converter and in the
other combustion-hydraulic one with hydrostatic reduction gear.
Keywords : ship propulsion systems, combustion-electric driving system, combustion-hydraulic driving system
BASIC DESIGN ASSUMPTIONS AND
CHOICE OF PROPULSION SYSTEM
The to-be-designed propulsion system is intended for the
passenger ship of the following technical parameters :
* overall length L = 56 m
* overall breadth B = 9 m
* draught d = 1 m
* ship displacement = 440 t for d = 1m
* assumed ship speed V = 14 km/h at d = 1m
* required power output P = 300 kW
The required very small draught of the ship is an impor-
tant limitation in searching for a suitable propulsion system.
Certainly it cannot be a propulsion system using cycloidal
propellers which are located under the ships hull. Ships driven
by water jets fairly well operate in shallow waters. However
such drive is unfavourable from the ecological point of view.
A large water stream sucked out from under ships bottom and
thrown overboard with a great velocity destroys bottom and
side structures of the waterway and biological live existing
there. In this case the factor has been deemed so important that
it was decided to exclude the water jet propulsion system from
further considerations.
Hence only the systems tted with screw propellers have
been taken into account. As the propeller is assumed to operate in
non-cavitating range, its appropriate diameter should be greater
than 1.4 m. In the case of two propellers used e.g. in Schottel
rotatable thrusters the diameter of each of them might be a little
smaller, equal to about 1.35 m.
Due to the small draught of the ship the above mentioned
values of propeller diameter are not acceptable. Therefore it
4 POLISH MARITIME RESEARCH, No 1/2006
1
3
8
6
7
10
2
9
4
5
was deemed necessary to apply a double-propeller propulsion
system. This way it would be possible to decrease the diameter of
the propellers to such an extent as to decrease its value to 0.83
0.85 m in the case of placing them in Korts nozzles, being still
large enough to transfer the assumed power. The outer diameter
of the nozzles would then exceed a little the draught of 1 m, but
if ships hull form is suitably corected it will not be a problem.
Such solution has many advantages. The double-propeller
propulsion system provides higher ships manoeuvrability and
reliability. Location of the propellers inside the nozzles signi-
cantly lowers risk of catching the propellers blade on the bottom
that usually results in a failure and necessity of replacement of
the propeller. Ducting the propellers also lowers unfavourable
inuence of screw race on the waterway bottom structure.
An additional improvement of reliability of the system and
its simplication can be obtain by applying the xed-pitch
propeller. However it requires to provide the system with
capability of changing magnitude and direction of rotational
speed of the propeller shaft, that can be realized in the simplest
way by a hydrostatic or electric transmission included in the
propulsion system.
Further advantages can be achieved by replacing the xed
nozzle with pivotable one, and even better by using a rotatable
thruster. This makes it possible to resign from applying the
traditional rudder and in consequence to signicantly decre-
ase gabarites and weight of the device and simultaneously to
improve ships manoeuvrability.
Taking into account the above presented factors one decided
to elaborate conceptual design projects of two solutions of
the propulsion system tted with rotatable thrusters, the most
technically justied in the opinion of these authors, namely :
4 combustion-electric one tted with typical asynchronous
squirrel-cage electric motors and frequency converters ma-
king stepless control of rotational speed of xed propeller
possible
4 combustion-hydraulic one tted with hydrostatic trans-
mission.
COMBUSTION-ELECTRIC
PROPULSION SYSTEM
The elaborated combustion-electric main propulsion system
of the ship in question is presented in Fig.1 and 2 in two versions
differing to each other mainly by size of electric motors and
way of their positioning.
Three electric generating sets were applied; each of them
consisted of a four-stroke combustion engine driving a three-
-phase synchronous generator.
The total power output of the generating sets fully covers
power demand for propelling and steering the ship. The output
of the third generating set sufces to cover the assumed power
demand of other consumers. So produced energy is delivered
to the main switchboard. From here its main part goes to the
frequency converters and next to the three-phase asynchronous
electric motors driving the xed screw propellers through the
toothed intersecting-axis gear.
In the rst version of the systemin question, shown in two
axonometric projections in Fig.1, have been applied the elec-
tric motors in vertical position, that made it possible to obtain
a modular construction of rotatable thruster, more compact
and having relatively small gabarites. The drive is transmitted
from the motors to the propeller shaft through the one-stage
toothed gear placed inside the pod (electric podded propulsor)
of rotatable thruster under water. Its reduction ratio is small
because of a limited size of the pod, that makes it necessary to
use a four-pole medium-speed electric motor.
a)
b)
The ship steering functions are realized by rotating the
column of rotatable thruster by an arbitrary angle around the
vertical axis. To this end were used two hydraulic motors of
a constant absorbing capacity, driving the column through
toothed gears. The motors are fed from a constant capacity
pump placed in the oil tank system.
A drawback of the solution is that the upper surface of
the electric motor sticks out a little over the rst deck of the
ship (not shown in the gure). Another unfavourable feature
is that the mass centre of electric motors is located high and
shifted aft.
In the second version shown in Fig.2. the electric motors
were placed in horizontal position. In consequence it was ne-
cessary to use a two-stage reduction gear of Z type. This way
Fig. 1. View of an example arrangement of the main components
of the combustion-electric propulsion system
tted with the electric motors in vertical position .
Notation : 1 electric generating set, 2 auxiliary electric generating set,
3 electric three-phase asynchronous cage motor driving the propeller,
4 frequency converter, 5 main switchboard, 6 rotatable thruster,
7 hydraulic unit for supplying hydraulic motors, 8 hydraulic motor
tted with planetary gear to drive the mechanism rotating the column
of rotatable thruster, 9 central outboard water fresh water cooler,
10 exhaust piping with silencers .
1
3
8
6
7
10
2
9
4
5
of the combustion-electric propulsion system tted
with the electric motors in horizontal position .
Notation : 1 electric generating set, 2 auxiliary electric generating set,
3 electric three-phase asynchronous cage motor driving the propeller,
4 frequency converter, 5 main switchboard, 6 rotatable thruster,
7 hydraulic unit for supplying hydraulic motors, 8 hydraulic motor
tted with planetary gear to drive the mechanism rotating the column
of rotatable thruster, 9 central outboard water fresh water cooler,
10 exhaust piping with silencers .
5
the total reduction ratio could be greater resulting in possible
application of two-pole electric motors of twice higher rota-
tional speed, smaller and somewhat lighter. However it should
be stressed that such solution is sometimes characterized by
signicant displacements of motors shaft axis against rotatable
thruster, caused by ships hull deformations in heavy weather
conditions. In order to eliminate the unfavourable inuence
of the deformations on operation of the toothed gear a exi-
ble connection of the motor shaft and gear shaft by means of
a Cardan coupling, was provided for.
Though such solution makes the construction more com-
plex and of greater gabarites, it does not lower weight of the
construction but only provides more favourable location of its
centre of gravity.
COMBUSTION-HYDRAULIC
PROPULSION SYSTEM
Fig.3 shows the propulsion system tted with hydrostatic
transmission, described in detail in [1]. To simplify the dra-
wing the oil piping and other auxiliary devices of power plant
have been omitted. The system is composed of two identical,
mutually independent sub-systems; each of them is driven by
the high-speed combustion engine (2). The engine directly
drives : the main oil pump of variable capacity (4) and through
the mechanical gear (3), the oil pump of constant capacity (not
shown in the gure) for driving the rotation mechanism of
the rotatable thruster, as well as the three-phase synchronous
electric generator (5).
The main oil pump (4) feeds the hydraulic motor of constant
absorbing capacity (7). The motor drives, through the toothed
bevel gear, the xed propeller located within pivotable nozzle
of the rotatable thruster (6). In this case the application of a -
xed propeller was justied by making it possible to steplessly
control speed and direction of rotation of the hydraulic motor.
This is realized by changing the capacity of the pump (4) and oil
pumping direction. The control system of the variable capacity
pump is fed from a small pump installed at the main pump.
The system was provided with one large oil tank (9) and
one central outboard water fresh water cooler (12) that
made it possible to reduce space of the ship power plant as well
as a number of its auxiliary devices. The electric generating
set (1) was applied to satisfy electric energy demand of other
consumers.
COMPARATIVE ANALYSIS
OF THE SYSTEMS
Final choice of the most favourable system is not an easy
task as it must take into account a broad range of various factors
including rst of all: manoeuvrability, reliability, initial and
operational costs, as well as area and space occupied by the
system, its mass, location of its centre of gravity etc.
In Tab. 1-3 and diagrams (Fig.4,5) are presented some im-
portant data concerning the features and costs of the considered
propulsion systems, which are supposed to make decision-
-taking easier.
As it follows from the given data both variants of combu-
stion-electric system are very similar. However some rather
important differences between them dealing with constructional
problems not clearly indicated in the tables, do exist. The ho-
rizontal position of electric motors, as compared with vertical
one, results in the necessity of application of the second stage
of the bevel gear, additional shaft and couplings between the
gear and motor, that makes the construction more complicated
and real initial cost of rotatable thruster greater. Moreover the
area occupied by the motor increases, however in that region
it probably would not be used for other purposes.
The vertical position of the electric motor is, in the solution,
a disadvantage resulting from a too-large height of the motor
xed on the rotatable thruster, not tting under the deck. How-
ever its importance may be effectively reduced by a suitable
arrangement of shipboard equipment in that region.
A greater number of even more important differences can
be revealed by comparing both systems to each other : com-
bustion-electric and combustion-hydraulic one.
1
3
8
6
7
10
2
9
4
5
11
12
of the combustion-hydraulic propulsion system .
Notation : 1- electric generating set, 2 combustion engine, 3 mechanical
gear, 4 the main pump unit to drive the propeller, 5 electric generator,
6 rotatable thruster, 7 hydraulic motor driving the propeller,
8 exhaust piping with silencers, 9 hydraulic oil supplying unit,
10 hydraulic motors to rotate the rotatable thruster around vertical axis,
11 electric switchboard, 12 central outboard water
fresh water cooler .
M
a
s
s

o
f

t
h
e

s
y
s
t
e
m

[
%
]
Fig. 4. Comparison of mass of the presented propulsion systems :
1 combustion-electric one with electric motors in vertical position,
2 combustion-electric one with electric motors in horizontal position,
3 combustion-hydraulic one .
Fig. 5. Comparison of initial costs of the presented propulsion systems :
1 combustion-electric one with electric motors in vertical position,
2 combustion-electric one with electric motors in horizontal position,
3 combustion-hydraulic one .
C
o
m
p
a
r
a
b
l
e

c
o
s
t

[
%
]
0
20
40
60
80
100
120
1 2 3
0
20
40
60
80
100
120
1 2 3
Name of unit Gabarites Number
Mass : of one
unit / total [kg]
Volume [m
3
] / area [m
2
]
Gross price (cost) :
of one unit / total [EURO]
Electric generating set of 240 kW
output : IVECO GE 8210SRM45
2737x1150x1371 2 2820 / 5640 8.63 / 6.30 57050 / 114100
2309x720x1280 1 1950 / 1950 2.13 / 1.66 37100 / 37100
Electric motor of 200 kW power :
EMIT SVgm 315 mL4
1210x600x800 2 1120 / 2240
1.16 / 0.96
not accounted for
in calculations *
6500 / 13000
Frequency converter :
Danfoss VLT
370x420x1600 2 200 / 400 0.50 / 0.31 18000 / 36000
Hydraulic oil tank unit 500x500x700 1 300 / 300 0.15 / 0.25
Rotatable thruster 2 910 / 1820 54000 / 108000
Total /12350 11.41 / 8.52 /~308200
Mass : of one
unit / total [kg]
Volume [m
3
] / area [m
2
]
2737x1150x1371 2 2820 / 5640 8.63 / 6.30 57050 / 114100
2309x720x1280 1 1950 / 1950 2.13 / 1.66 37100 / 37100
Electric motor of 200 kW power :
EMIT Sgm 315 mL2
1200x1000x615 2 1100 / 2200
1.48 / 2.4
not accounted for
in calculations *
18000 / 36000
Frequency converter :
Danfoss VLT
370x420x1600 2 200 / 400 0.50 / 0.31 5200 / 10400
Hydraulic oil tank unit 500x500x800 1 300 / 300 0.20 / 0.25
Total /12450 11.41 / 8.52 /~317600
Mass : of one
unit / total [kg]
Volume [m
3
] / area [m
2
]
Combustion engine of 220 kW
output : IVECO CURSOR 300
1770x935x1030 2 900 / 1800 3.41 / 3.31 24500 / 49000
output : IVECO GE8210SRM36
2975x1110x1940 1 2520 /2520 6.41 / 3.30 47160 / 47160
Hydraulic pump of 252 kW maximum
output : Rexroth A4VSG 180
350x300x220 2 114 / 228 0.05 / 0.21 12000 / 24000
Hydraulic motor of 220 kW maximum
output : Rexroth A2FM 250
224x280x250 2 73 / 146
0.03 / 0.13
not accounted for
in calculations *
5368 / 10736
Electric generator of 40 kW output :
Leroy Somer 42.2VL8
615x450x450 2 165 / 330 0.25 / 0.55 3000 / 6000
Hydraulic oil tank unit 1000x1000x1300 1 1050 / 1050 1.23/ 1
Total / 7874 11.31 / 8.37 /~244900
Tab. 3. Parameters and prices (costs) of the units of the combustion-hydraulic propulsion system with hydrostatic transmission .
Tab. 2. Parameters and prices (costs) of the units of the combustion-electric propulsion system with electric motors located in horizontal position .
Tab. 1. Parameters and prices (costs) of the units of the combustion-electric propulsion system with electric motors located in vertical position .
* This unit does not occupy any useful space of the power plant
7
On 20-22 May 2005, under this slogan, Faculty of
Mechanical Engineering, Gdask University of Tech-
nology, organized the international symposium which
was a continuation of meetings of the Facultys staff and
representatives of German partnership scientic research
centres.
25 years ago such meetings have started from the con-
tact with Hochschule in Bremen and in the course of time
the number of German participants has increased.
As Poland entered the European Union, the following
program assumptions were announced for the Symposium
of 2005 :
O presentation of advances and latest achievements in
mechanical engineering as to enable common appli-
cations for European grants
O experience and information transfer in the eld of
implementation of European curricula in academic
education at technical universities
O discussing problems of industrial implementation of
research results
O establishing new personal contacts and strengthening
existing friendly relations between scientists.
and also scientic workers fromFrance and Great Britain
took part in the Symposium.
The Symposium program contained presentation and
discussion of 53 papers out of which 29 were submitted
by the organizers, 8 by scientic workers from Stralsund
University, and the rest by representatives of other scien-
tic centres : 8 German, 2 French, and 1 British, as well
as 2 Polish ones (Czstochowa University of Technology
and Institute of Fluid Flow Machinery, Gdask).
As far as the broadly understood ships manoeuvrability is
concerned the propulsion system with hydrostatic transmission
shows more advantages out of which the following are most
important :
better protection of the propulsion system against overlo-
ading that results in much higher reliability and durability
of its units especially the bevel gear
more accurately and faster realized manoeuvers that mainly
results from many times smaller inertia moments of the
hydraulic motors as compared with those of electric ones.
Successive advantages of the combustion-hydraulic system
are its smaller weight and gabarites. They rst of all result from
several times smaller mass of the hydraulic motors and space
occupied by them against those of the electric motors of the
same power. As results from the data included in the tables the
masses differ to each other almost fteen times. Another source
of the merits is the application of the high-speed combustion
engines for driving the hydraulic pumps, having rotational
speed much higher than that of the electric generating sets used
in the combustion-electric propulsion system. The features are
especially favourable in the case of small vessels especially
those intended for sailing in shallow waters.
Another important advantage of the system tted with hy-
drostatic transmission is its smaller initial cost amounting to
less than 77% of that of the remaining systems in question.
An important drawback of the considered system is its
higher operational costs. They mainly result from a lower
efciency of the system. From the so far performed analyses
it results that the efciency of the combustion-electric system
is by about 5% higher than that of the combustion-hydraulic
system, at rated values of their operational parameters. For this
reason in the combustion-hydraulic system a somewhat higher
total power output of combustion engines has been provided
for. The need of periodical change of oil and ltering cartridges
additionally rises operational costs.
FINAL REMARKS
All the three presented design variants of the ship propulsion
systems satisfy the assumptions enumerated in Introduction
and each of them could be applied to the designed ship. To
choose the most favourable one out them is not an easy task.
However these authors are convinced that the above presented
analysis of basic features and costs of each of the systems cer-
tainly may help the principal designer of the ship in making
a proper choice.
BIBLIOGRAPHY
1. Cz. Dymarski, G. Skorek: : Energy balance of conceptual design
of propulsion system with hydrostatic transmission for inland
and coastal vessel(in Polish). Faculty of Ocean Eng. and Ship
Techn., Gdask University of Technology. Research Report
No 154/E/2004
2. Cz. Dymarski; R. Rolbiecki Conceptual design of propulsion
and control system with hydrostatic transmission for inland
vessel(in Polish). Faculty of Ocean Eng. and Ship Techn.,
Gdask University of Technology. Research Report
No 164/E/2004
3. Cz. Dymarski: An azimuthing combustion-hydraulic propulsion
system for inland vessel. Marine Technology Transaction
(Technika Morska), Vol. 16, 2005
4. Information pamphlets and offer materials of the following
companies: KaMeWa, Lips, Schottel, Ulstein, Wartsila. Iveco
Motors, Bosch Rexroth, Emit, Schottel, Leroy Some .
CONTACT WITH THE AUTHORS
Assoc. Prof. Czesaw Dymarski
Rafa Rolbiecki, M.Sc.,Eng.
Faculty of Ocean Engineering
and Ship Technology,
Narutowicza 11/12
Research Education Technology
Numerical simulation
of viscid ow around hydrofoil
Pawe Dymarski
Ship Design and Research Centre Stock Company, Gdask
ABSTRACT
This paper presents results of application of a viscid uid model to calculate ow around
a nite-span hydrofoil. The presented calculations were performed with the use of SOLAGA
software developed by this author. The used theoretical model of viscid liquid motion was
described, consisted of averaged liquid motion equations and Spalart-Allmaras one-equa-
tion model of turbulence. Also, a numerical model based on the nite volume method was
shortly presented. Calculation results were shown in the form of distribution diagrams of
pressures, longitudinal component of velocity and longitudinal component of rotation ve-
locity. Additionally, the presented characteristics of the lifting force coefcient and drag force coefcient
were compared with experimental data.
Keywords : viscid ow around hydrofoil, Spalart-Allmaras model, Finite Volume Method, SOLAGA
INTRODUCTION
Hydromechanical calculations of ship screw propellers car-
ried out by Polish research centres (Ship Design and Research
Centre Stock Company, Institute of Fluid Flow Machinery
- Polish Academy of Sciences, and Gdask University of Tech-
nology) are mainly based on potential motion of liquid, i.e. such
as used in vortex methods or surface panel methods. In recent
years in shipbuilding have been often and often applied pro-
grams dealing with viscid liquid ow. The programs are based
on the Reynolds Averaged Navier-Stokes Equation (RANSE)
and turbulence model equations. Solvers based on the RANSE
methods have been for years applied to calculate ship wake,
and quite recently they have been used as universal programs
for solving any hydromechanical problem.
The project under development, whose partial results have
been presented in this paper, is aimed at elaboration of an al-
gorithm and computer program for calculating the viscid ow
around ship propellers and hydrofoils. This author is convinced
that development of a computer program specialized in cal-
culating objects of one kind will make it possible to obtain -
- in the future - better results against those achievable from the
universal solvers widely applied (mainly abroad).
In this paper are presented calculation results obtai-
ned by means of the SOLAGA computer software under
development within the frame of the project in question.
The test calculations were performed for a rectangular hydrofoil
of NACA66-9 prole and the aspect ratio = 6.
VISCID FLUID MOTION EQUATIONS
The closed system of uid motion equations is based on two
principles : the principle of mass conservation and principle of
conservation of momentum.
The equation of mass conservation formulated
for inviscid uid is as follows :
(1)
And the equation of conservation
of momentum has the following form :
(2)
where :
(3)
is the strain - rate tensor and :
p - pressure
- density
- dynamic viscosity coefcient
x
i
, u
i
- components of location vector
and velocity vector, respectively.

0
x
u
i
i
=

i j
j i j
i
i
i
s 2
x x
p
x
u
u
t
u

|
|
.
|
\
|
c
c
+
c
c
=
i
i
j
j
i j
x
u
x
u
2
1
s
9
The above given equations fully describing the behaviour of
Newtonian incompressible liquid are applicable in calculating
the ows characterized by small values of Reynolds number,
i.e. up to about 10
3
10
4
.
In the case of ows of a greater Reynolds number, to apply
the equations is not possible due to the scale of phenomena
occurring in the uid.
In the engineering problems associated with owof water,
Reynolds numbers as usual greatly exceed the value of 10
4
. In
ship hydromechanics its value is as a rule contained within the
interval of 10
6
10
8
, and in this connection instead of Eqs (1)
and (2) their averaged equivalents are used.
The instantaneous velocity vector u
i
(x,t) can be presented
as a sum of the average velocity U
i
(x) and the uctuation
' u
i
(x,t) :
(4)
In a similar way the instantaneous pressure
eld p(x,t) can be described :
(5)
Applying the above notation used to the set of Eqs (1),(2)
one obtains the averaged equations of motion [6] :
(6)
(7)
In the averaged momentum conservation equation the term
i j
' u ' u (called Reynolds stress tensor) appears. In order to
compute the tensor it is necessary to introduce additional
equations which are called turbulence model. A turbulence
model is selected respective to a considered kind of ow and
an assumed calculation accuracy; calculation time is also one
of the important factors - the more complex model the longer
time of calculations.
In the presented calculation program the Spalart-Allmaras
one-equation model was used, introduced for calculating hy-
drofoils.
The turbulent stress tensor is calculated on the basis of Bous-
sinesqs hypothesis which assumes that the Reynolds stresses
can be expressed by means of the turbulent viscosity and mean
strain - rate tensor :
(8)
The kinematic turbulent viscosity
T
is calculated from the formula :
(9)
And, the modied turbulent viscosity
is calculated with the use of the transport equation :
(10)
The values of the models coefcients
and the relations between the coefcients are given below :
(11)
( ) ( ) ( ) t , ' u U t , u
i i i
x x x + =

( ) ( ) ( ) t , ' p P t , p x x x + =

0
x
U
i
i
=

( )
i j i j
j i j
i
i
i
' u ' u S 2
x x
P
x
U
U
t
U

|
|
.
|
\
|
c
c
+
c
c
v = = t
i
j
j
i
T j i ij
x
U
x
U
' u ' u
~
1 v T
f

\
|
v ( )
k k
2 b
k k
2
w 1 w SA 1 b
j
j
x
~
x
~
c
x
~
~
x
1
d
~
f c
~
S
~
c
x
~
U
t
~
c
v c
c
v c
o
+
(
c
v c
+ v
c
c
o
+
+
|
.
|
v
v =
c
v c
+
c
v c
3 / 2 , 1 . 7 c , 622 . 0 c , 1355 . 0 c
1 v 2 b 1 b
= = = =
(12)
(13)
(14)
(15)
where :
ij rotation tensor
d distance to the nearest surface of the ow-around object .
(16)
The assumed initial and boundary conditions
The value of the modied turbulent viscosity equal to 0.1
(where : coefcient of molecular kinematic viscosity) was
assumed at inlet. On the hydrofoil surface = 0 is assumed
as the turbulent viscosity in the ow close to wall amounts to
zero. [2].
NUMERICAL MODEL
The below presented equations are solved in a discrete
way they are transformed into a set of algebraic equations
formulated for each node of calculation grid. Depending on
specicity of a considered problem one can use one of the
following space discretization method :
In the SOLAGA software the discretization is performed by
means of the FVM which provides a great freedom in selecting
a kind of calculation grid, that makes it possible to calculate
ow around objects of complex geometry.
Discretization of computation space.
The Finite Volume Method (FVM)
The initial point to formulate the FV Method is the integral
form of conservation equations, namely a considered calcula-
tion space is divided into a nite number of control volumes and
for each of them relevant behaviour equations are formulated.
Eq. (17) shows the general form of the conservation equation
of the scalar quantity , given in the integral form :
(17)
where :
V velocity vector
eld scalar function
n vector normal to the control element surface S
diffusivity
q
source of the quantity .

41 . 0 , 2 c , 3 . 0 c
c 1 c
c
3 w 2 w
2 b
2
1 b
1 w
= = =

+
+
=
6 / 1
6
3 w
6
6
3 w
w
c g
c 1
g f

+
+
=
1 v
2 v
3
1 v
3
3
1 v
f 1
1 f ,
c
f
+
=
+

=
( )
2 2
SA
6
2 w
d S
~
~
r , r r c r g ,
~
= + =
=
ij ij SA 2 v
2 2
SA SA
2 S , f
d
~
S S
~
=
+ =
~
~
Finite Differences Method (FDM) [4]
Finite Element Method (FEM) [3]
Finite Volume Method (FVM) [4].
=
i
j
j
i
ij
x
U
x
U
2
1

+ = d q S d grad S d
S S
n n V
The surface and volume integrals appearing in the above
given equation are calculated by means of approximate me-
thods. Values of the eld function between computation nodes
are obtained by using interpolation.
Calculation of surface and volume integrals
The surface and volume integrals appearing in Eq. (17)
are calculated by means of the below described quadratures.
The integral of the function f on the surface S
can be developed as follows (Fig. 1) :
(18)
where :
S
k
- surface of the wall with the index k (k = w, e, s, n, b, t).
The integral of the function f on the single wall S
e
is appro-
ximately calculated by applying the central point method :
(19)
where : f
e
- value of the integrand in the central
point e of the surface S
e
.
The volume integrals are calculated in an analogical way;
the integral of the eld function q over the volume can be
calculated as follows :
(20)
where : q
P
- value of the function q in the central point P.
Field function values in the central points of control vol-
umes (nodes) are obtained directly from solving the set of
equations, whereas eld function values on the control volume
surface are calculated by means of appropriate interpolation
schemes.
Interpolation schemes
Carrying out calculations with the use of SOLAGA software
one can apply two kinds of interpolation schemes :
Fig. 1. Control element of 3D grid containing the node P. The nodes :
W, E, S, N, B, T of the neighbouring elements and the central points
of the element walls w, e, s, n, b, t are also depicted.
e e
S
S f S d f
e
P P
q d q Q
In the upwind method, transport of the medium described
by the function is assumed to be realized mainly by convec-
tion value of the function in the central point of the elements
wall is equal to that in the neighbouring node located upwind
the medium ow. Value of the function in the point e (Fig.2)
is calculated by using the following formula :
(21)
The upwind scheme is unconditionally stable and does not
lead to oscillating solutions. Its drawback consists in a large
numerical diffusion and low accuracy (of 1st order).
More accurate results can be obtained by using the linear
interpolation scheme (CDS) based on the assumption that the
function between the points P and E varies linearly (Fig.2 b).
Value of the function in the point e is calculated as follows :
(22)
where :
e
- linear interpolation coefcient calculated
by means of the following formula :
(23)
Eq. (22) is of the degree of 2nd-order accuracy. The CDS
scheme is characterized by a much lower numerical diffusion
than that of the upwind one. The drawback of the CDS scheme
is its lower stability and susceptibility to generate oscillating
solutions.
* Upwind interpolation (UDS)
* Linear interpolation (CDS).
( )
e P e E CDS , e
1 + =
P E
P e
e
x x
x x

=
k
k
S S
S d f S d f
Fig. 2.a) Cross-section through 3D calculation grid. Note : control elements
with the nodes P and E and the wall located between them and containing
the central point e, are marked grey. Value of the function in this point
is obtained by using interpolation.
In Fig. 2.b) are shown principles of operation of the interpolation schemes:
UDS upwind interpolation, CDS linear interpolation.
a)
b)
( )
( )
<
>
=
0 if
0 if
E
P
UDS , e
e
e
n V
n V
11
In order to avoid the solution instability the both schemes
are connected together by introducing the so called blending
factor B 0 ; 1. The interpolated value of the function in
the point e is hence determined as follows :
(24)
where :
e
,UDS - value calculated by means of the UDS scheme

e
,CDS - value calculated by means of the CDS scheme.
This approach makes it possible to achieve a stable solution
at maintaining a relatively high degree of calculation accuracy.
Numerical methods
of solving non-stationary problems
In the case when considered problems have a non-stationary
character it is necessary to apply the methods analogous to
those used in solving initial problems for ordinary differential
equations.
The problems associated with liquid oware as a rule cha-
racterized by time-variability in spite of stationary boundary
conditions, in the cases application of a non-stationary model
is recommended. Such approach is especially justied in the
case of calculation of rather non-streamlined objects where
ow separation may occur.
Methods of solving initial problems
of ordinary differential equations
After spatial discretization of the partial differential equ-
ation the set of ordinary differential equations is obtained in
the following form :
(25)
The simplest way of solving the above given equation is
its integration within consecutive time intervals to obtain the
solution for the successive time values t
1
, t
2
, t
3

(26)
The above given expression is accurate. Numerical methods
for determining the quantity
+1 n
are classied with accounting

for a way of calculation of the right-hand side integral. In the pre-
sented software two integration methods are applied : the back-
ward (implicit) Euler method and three time level method.
In the Euler method a value of the integrand within the whole
interval t t
n
, t
n+1
is assumed equal to F
[t
n+1
,
(t
n+1
)] , see
Fig.3a. Elements of the vector
in the successive time-step

are calculated as follows :
(27)
The three time level method consists in approximating the
course of the function F
[t
n+1
,
(t
n+1
)] by means of 2nd order
parabola crossing the points t
n-1
, t
n
, t
n+1
, see Fig.3b. The
formula of the method is as follows :
(28)
Both the mentioned methods are of implicit kind, i.e. that
knowledge of values of its elements is necessary to calculate
the vector
+1 n
. As a result of the application of one of the

above described methods a set of m - equations is formed in
every time step, where m stands for a number of calculation
nodes. The equation set is solved in an iterative way by using
the methods described below.
Methods of solving the sets of equations
As a result of spatial discretization of motion equations
a set of algebraic equations formulated for each of the control
volumes is obtained. The set, when linearized, is solved in an
iterative way. In the software in question the algorithms based
on the method of conjugate gradients are used : ICCG algori-
thm serves for solving the symmetrical matrices obtained from
discretization of the Poisson equation (calculation of pressure
corrections), and the unsymmetrical matrices are solved by
means of Bi-CGSTAB algorithm. The above mentioned algo-
rithms are described in [4].
TESTING CALCULATIONS
Computational data
The geometry of hydrofoil
Chord : c = 0.2 m
Span : w = 1.2 m
Prole : NACA66-9
Angles of attack 0, 2, 4, 6, 8, 10, 12 degrees
Inow velocity V = 5 m/s
Model details
Turbulence model
Spalart-Allmaras
(see Eqs 8 16)
Interpolation scheme
UDS and CDS
(blending factor B = 0.5)
Time integral
approximation scheme
Implicit Euler (1
st
order)
Time step 0.002 0.01 sec
Number of control volumes 600 000
( )
CDS , e UDS , e e
B B 1 + =
( )
( ) [
[
( )
0
0
t , t , t F
t d
t d
= =
( ) t , t F
1 n
1 n
n 1 n
+ =
+
+
+

( ) t , t F
3
2
3
1
3
4 1 n
1 n
1 n n 1 n
+ =
+
+
+

Fig. 3. Numerical integration of equations for non-stationary problems :
a) implicit Euler method, b) three time level method .
a)
b)
( )
( ) [ ]

+ +
= =
+
1 n
n
1 n
n
t
t
n 1 n
t
t
t d t , t F t d
t d
t d


Results
Maps of pressure distribution on the hydrofoil plane of symmetry for various angles of attack are shown in Fig.4.
And distribution of rotation in the wake behind the hydrofoil is shown in Fig.5. Distribution of the longitudinal
component of rotation is presented in Fig.6.
Fig. 5. Distribution of longitudinal velocity component just over hydrofoils
surface, in its plane of symmetry, and on the plane normal to ow direction,
distant by one chord length behind trailing edge of the hydrofoil .
Fig. 4. Pressure distributions on surface of hydrofoil
and its plane of symmetry .
2
4
8
12
2
4
8
12
13
CONCLUSIONS
The computational results presented in this paper
may be concluded as follows :
O The calculated pressure distribution around hydrofoil is
qualitatively similar to that obtained from experiment.
O The calculated distribution of longitudinal component of
rotation behind the hydrofoil is similar to the rotation elds
obtained from the experiments [5]. The tip vortex core is
clearly visible. The strength of vorticity increases up along
with the angle of attack.
O The coefcients of lifting force and drag force are in a good
conformity with the experimental data especially in the ran-
ge of angle of attack from 0 to 6 degrees. The critical point
of the lifting force coefcient characteristics is predicted
with the accuracy of about 1 2 degrees.
Acknowledgement
The research presented in this paper has been nancially
supported by the Polish Ministry of Science and Information
Society Technologies, (Grant No. 5 T12C 012 22). The author
would like to express his gratitude for this support.
NOMENCLATURE
C
D
- drag force coefcient
C
L
- lifting force coefcient
d - distance to the nearest surface of the ow - around object
f - eld function which determines a given (diffusional and/or
convectional) ow rate
F
- vector of integrand functions of a set of ordinary

differential equations (given in vectorial form)
n - vector normal to the surface S
p - instantaneous pressure
p' - pressure uctuation
P - time-averaged pressure
q - function which determines eld source
Q
P
- total source in control element containing the central point P
S - control surface (which limits control element)
S
ji
- time-averaged strain - rate tensor of deformations
t - time
t
n
- t value in n-th step of calculations
u
i
- instantaneous velocity component (in Cartesian notation)
u'
i
- uctuation of velocity component
U
i
, U
j
- time-averaged velocity components
V - velocity vector, see Eq.(17)
x - location vector
2
4
8
12
Fig. 6. Distribution of longitudinal rotational velocity component on the
plane normal to ow direction, distant by one chord length behind trailing
edge of the hydrofoil. The multi-colour spot on the right-hand side shows
distribution of rotation in the tip vortex core .
The computational characteristics of the coefcients of
lifting force and drag force are compared with the experimental
data [1] in Fig.7.
[deg]
+
+
+
+
+
+
X
X
X
X
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
C
,
C
[
-
]
L
D
[
-
]
0 2 4 6 8 10 12 14
C
C
(C )
(C )
L
D
L
D
ex p
ex p
X
+
Fig. 7. Coefcients of the lifting force C
L
and the drag force C
D
in function
of the angle of attack. The experimental data are in accordance with [1] .
x
i
- component of location vector
- diffusion coefcient of the quantity
- dynamic viscosity coefcient of a liquid
- kinematic viscosity coefcient of a liquid
- turbulence model coefcient (turbulent modied viscosity),
see Eq.(9)
T
- kinematic turbulent viscosity coefcient of a liquid
- density
- eld scalar function
- vector of function values in nodes of calculation grid

(for a discrete form of function)
e
- value of function in the point e
E
- value of function in the point E
e,UDS
- value of
e
approximated by using UDS scheme
e,CDS
- value of
e
approximated by using CDS scheme
p
- value of function in the point P
- control element volume.
Acronyms
Bi-CGSTAB - BiConjugate Gradient Stabilized method
ICCG - Incomplete Cholesky Conjugate Gradient method
BIBLIOGRAPHY
1. Abbott I.H., von Doenhoff A.E. : Theory of Wing Sections.
Dover Publications inc. New York, 1959
~
2. Blazek J. : Computational Fluid Dynamics: Principles and
Applications. ELSEVIER. 2001
3. Chung T.J : Computational Fluid Dynamics. Cambridge
University Press. 2002
4. Ferziger J.H., Peric M. : Computational Methods for Fluid
Dynamics. Springer. Berlin, 1999
5. Koronowicz T. : Rationality of Solving 3D Circulation Problems
Exclusively with the Use of Navier-Stokes Equation. TASK
quarterly No 1/2002
6. Wilcox D.C.: Turbulence Modeling for CFD, DCW Industries.
2002 .

CONTACT WITH THE AUTHOR
Pawe Dymarski, M.Sc.,Eng.
Ship Hydromechanics Division,
Research and Development Department,
Ship Design and Research Centre Stock Company
Szczeciska 65

The Congress in which representatives participated of
universities and scientic research centres as well as auto-
motive industry rms from9 European countries, Japan and
USA, took place in Bielsko-Biaa, a town in the mountainous,
south-western region of Poland.
Its program contained 3 plenary sessions with 4 papers
presented during each of them, under the slogans :
Automotive/Engine Technology Development (4 papers)
Advanced Engine Design & Performance (8 papers)
and 5 technical sessions covering
the following groups of topics :
The third part was the poster session devoted to presenta-
tion of 42 different elaborations. Among so many different
papers 9 of them dealt with ship engines, namely :
* Ways to improve operational properties of lorry engines
by Janusz Mysowski and Jaromir Mysowski (Technical
University of Szczecin)
* Universal research test of toxic exhausts from ship piston
engines by L. Piaseczny and T. Kniaziewicz (Polish
Naval University)
* Possibilities of recognizing faults in gas turbine engines
on the basis of simulation of transitory processes by A.
Adamkiewicz (Polish Naval University) and M. Dzida
(Gdask University of Technology)
* Analysis of vibration parameters of marine gas turbine
engines by A. Grzdziela (Polish Naval University)
* Diagnostic examination of marine engines in the Polish
Navy by Z. Korczewski (Polish Naval University)
* Construction of modern valve train mechanism for mari-
ne diesel engines by T. Lus (Polish Naval University)
* Investigations of carbon deposits on injector nozzles of
marine diesel engines by J. Monieta and P. Wjcikow-
ski (Maritime University of Szczecin)
* Research on inuence of delivery of water to cylinders
on combustion process parameters and exhaust toxicity
of internal combustion engines by L. Piaseczny and R.
Zadrg (Polish Naval University)
* Approximation of the cylinder compression pressure of
the marine engine by means of a multi-parameter model
by S. Polanowski (Polish Naval University)
Also, the panel discussion on
Development trends of internal combustion engines
was carried out within the frame of the Congress.
PTNSS Congress 2005
On 2528 September 2005 was held 1
st
International Congress on Combustion Engines
organized by Polish Scientic Society of Combustion Engines (PTNSS), under the slogan :
The development of combustion engines
+ Engine testing (28 papers)
+ Emission (17 papers)
+ Fuel injection (9 papers)
+ Combustion process (9 papers)
+ Modeling (8 papers)
+ Alternative fuels (6 papers)
+ Various subjects (4 papers)
15
POLISH MARITIME RESEARCH, No 1/2006 37
On 9-13 May 2005 under this heading was held 4
th
International Scientic Technical Conference organized
by the Faculty of Ocean Engineering & Ship Technology,
Gdask University of Technology, together with MAN-B&W
DIESEL A/S.
In accordance with the Conference slogan :
Keeping diesel engines and gas turbines
in movement with regard
to environmental protection
the Conference aim was to make science-practice relations
closer and to partricipate in creation of a forum for exchange
cognitive and utilitarian information on designing, manu-
facturing and operating the self-ignition engines and gas
turbines as well as machines and other devices necessary to
maintain the engines running, with special accounting for :
their power and pro-ecological qualities as well as durability,
reliability, diagnostics and operational safety.
The programs intentions were realized by 50 papers
published in the Conference proceedings, 25 of which were
presented and 20 were demonstrated during two poster ses-
sions, as well as by 6 presentations performed by represen-
tatives of the companies working a.o. in the eld of marine
engineering, namely :
Marine diesel engines and catalytic nes a new stan-
dard to ensure safe operation by G. Astroem (ALFA
LAVAL)
Activity prole by . Rzga (PBP ENAMOR Ltd)
ME engine concept in front of environment protection
demands of today and in the future by S. Henningsen
(MAN-B&W DIESEL A/S)
The latest MAN-B&W DIESEL design achievements on
the base of S65ME-C engine concept by A. Oestergaard
(MAN-B&W DIESEL A/S)
Application of portable electronic indicators in operation
of self-ignition engines by L. Tomczak (UNITEST)
Application of 3D visualization in marine training soft-
ware by L. Tomczak (UNITEST).
The course of the Conference was very attractive as
four rst sessions including one poster session were held in
a hotel at Midzyzdroje, a health resort on the coast of the
Baltic Sea, and the 5th session on board a ferry calling at
Copenhagen where the 6th and 7th sessions were arranged
in the MAN-B&W DIESEL headquarter.
There was also an occasion to visit a very interesting
museum of the company and to be acquainted with running
stand tests on the original 4T50MX two-stroke engine. Next
day, after return to Midzyzdroje the last four sessions had
place including one poster session.
Worthmentioning that the Conferences in question have
been characterized traditionally by a very valuable consi-
stence of theoretical knowledge and research and production
practice.
Representatives of 16 scientic research and design cen-
tres took part in the preparation of papers for the Conference.
The greatest share in this had scientic workers of Gdask
University of Technology, Polish Naval University, Gdynia
Maritime University, Pozna University of Technology,
Airforce Technical Institute and Maritime University of
Szczecin, who participated in the preparation of 12, 10, 7,
6, 4, and 3 papers, respectively.
EXPLO-DIESEL
&
GAS TURBINE05
Copenhagen R&D center

MAN - B&W DIESEL - Copenhagen R&D centre
On the application of the articial neural
network method to a neural simulator
of steam turbine power plant
Justyna lzak-ona
ABSTRACT
In the paper a neural simulator of steam power unit is presented as an example of appli-
cation of articial neural networks (ANN) for modeling complex technical objects. A set of
one-directional back-propagation networks was applied to simulate distribution of main
steam ow parameters in the cycles crucial points for a broad range of loading. A very
good accuracy and short computation time was obtained. The advantages make the simula-
tor useful for on-line diagnostic applications where short response time is very important.
The most important features of the simulator, main phases of its elaboration and a certain
amount of experience gained from solving the task was presented to make the practical application of the
method in question more familiar.
Keywords : neural modeling and simulating, turbine power plants, on-line diagnostics
INTRODUCTION
Steam power unit or ship power plant is a very complex
object. To know its technical state is crucial for carrying out
its operation in an optimum way, in which diagnostics is of
a great importance.
Today, apart from safety, the taking care of operational pro-
cess quality to obtain long-term reduction of cost has become
a priority. It consists in expanding times between overhauls at
simultaneous maintaining the efciency of devices on a con-
stantly good level. This is on-line diagnostics which makes
continuous controlling the technical state of objects under
operation possible. As it brings large economical prots the
diagnostics becomes more and more important and its dyna-
mical development can be thus explained.
Therefore is needed a device which would be able to accu-
rately determined operational parameters of a given object so
fast as to make it possible to compare them with current ones.
To achieve that determination time of a correct operational
standard should be of the order of milliseconds (resulting
from sampling frequency of measuring systems). Heat ow
diagnostics of steam power units is based on advanced analy-
tical models. However because of their long computation time
they can be used in off-line mode only. It means that periodic
control of technical state of an object can be performed on the
basis of earlier collected data.
The neural simulator operates as a black box and articial
neurons acts here instead of sophisticated models. Its response
process consists in simple mathematical operations. Due to this
fact a neural simulator is more primitive than an analytical one
but it provides standard operational parameters of a given object
very fast and with good accuracy that justies its application
to on-line diagnostics.
PHASES OF ELABORATION
OF THE NEURAL SIMULATOR
Choice of a simulated object
A standard steam power unit of 200 MW output tted with
a modernized TK 200 turbine was selected as the object to be
simulated (Fig.1). Such selection has been justied by the wide
application of units of the kind in Polish electro-energy system.
Training data acquisition
In the ANN method a fundamental thing is to have an
appropriate set of training data as the rules written in neural
model structure are generated on their basis. During training
the network nds only relations between a given input and
output, contrary to an analytical model elaborated on the
basis of universal laws of mathematics and physics where
experimental data serve only to control if theoretical laws are
in compliance with reality.
Hence to apply the ANN method it is necessary to collect
in advance a huge amount of experimental data for training
the network.
For lack of operational data of a real steam power unit this
author made use of DIAGAR software [2] which served as
a source of data for elaborating the neural simulator in question.
Such situation where an analytical simulator provides training
data is very advantageous as it makes it possible to generate an
almost arbitrary set of images for training the network.
Fig.2 shows schematic diagram of computations of the
object taken into account in DIAGAR software. The main
steam jet is marked red and regenerative steam extractions are
signed with Roman numerals.
17
Fig. 1. Simplied schematic diagram of the steam power plant, where : B Boiler, C - Constant pressure condensers, D Degasier, DP - Drip pumps,
G - Electric generator, HP - High Pressure unit of condensing turbine, HRH - HP regenerative heaters, LP - Low Pressure unit of condensing turbine,
MP - Medium Pressure unit of condensing turbine, RH - LP regenerative heaters, SC - Steam cooler, SI - Steam injectors,
VC - Cooler of vapours from stufng boxes, WP - Water supply pump, WT - Water supply tank .
Preparation of the training data set
The training data set consists of independent and dependent
operational parameters of the power unit [1], namely :
Set of independent parameters :
which dene loading state of the steam cycle, see Tab.1.
Set of dependent parameters :
The heat owparameters of working mediumare determi-
ned in 176 points of the cycle. As a result the data set covered
a wide range of the power units work, and amounted to 6300
combinations dening various loading states.
+ Turbine sets power output
+ Fresh steam pressure
+ Fresh steam temperature
+ Superheated steam temperature
+ Pressure in condenser,
4 Mass ow rate (m)
4 Pressure (p)
4 Temperature (t)
4 Enthalpy (h).
Tab. 1. Set of the parameters dening loading states of the power unit [5]
(Parameters of rated working state are marked red.) .
The task consisted in training the network in order to de-
termine a set of diagnostic (dependent) parameters in response
to a given set of independent operational parameters of the
power unit.

Set of independent operational parameters of the unit
Power
Fresh
steam
pressure
Fresh steam
temperature
Superheated
steam
temperature
Pressure in
condenser
Number of
combinations
N [MW] p
o
[bar] T
1
[C] T
2
[C] p
k
[bar] 6300
120 110 510 510 0.04
140 120 520 520 0.05
160 130 530 530 0.06
180 140 540 540 0.07
200 150 550 550 0.08
560 560 0.09
0.10
Choice of structural arrangement
of the network and algorithm of its training
During the last 20 years neural networks have been deve-
loped very dynamically. Their structures have been improved
and new algorithms elaborated. However choice of optimum
parameters has still remained a time-consuming process as it is
realized with the use of trial-and-error method. The structures
and algorithms of training the networks of which the neural
simulator is consisted, were preliminarily selected on the basis
of theoretical knowledge [3,4] and the published comparative
analyses [3].
The most important element of the selection process was
the testing of effectiveness of particular structures in solving
the task in question.
Finally, the structure having two processing layers : a hidden
layer of activating sigmoidal functions and an output layer of
linear functions, was selected. The series of trial trainings [5]
revealed that the most effective training algorithm for the task
in question is that of Levenberg-Marquardt (LM). This is one
of the fast convergent algorithms which not only converge after
a small number of training iterations but also are much superior
regarding networks response accuracy than other algorithms.
However the advantages are achieved at expense of high re-
quirements for RAM memory of used computer.
Training process of the network
To train the one-direction network under control the error-
-back-propagation algorithm is usually applied. The error value
determined in one iteration serves as the basis to correct weights
and thresholds for the next iteration. This way a continuous
improvement of networks response quality is achieved. The
training terminates when response accuracy determined by
comparing the response with an assumed standard, is satis-
factory. However many problems have been met in practice,
namely :
Too small capacity of RAM memory
of the applied computer
To obtain satisfactory accuracy of the simulator a very
large set of training data was required. It resulted in very
large dimensions of the training matrices (5 x 6300 and
17 x 6300). The next element was a large capacity of me-
mory demanded by the LM algorithm. A remedy for such
situation was to limit the number of simulated points of the
Fig. 3. Schematic diagram of the networks structure [5],
where :
a1 - response matrix of 1st neural layer (hidden one)
a2 - response matrix of 2nd neural layer (network response)
b1 - vector of weight coefcients for 1st neural layer
b2 - vector of weight coefcients for 2nd neural layer
n1 - matrix of neurons in 1st layer
n2 - matrix of neurons in 2nd layer
P - matrix of network training images
Q - number of network training vectors (images), Q = 6300
R - training vector of 5 elements, R=5
S1 - number of neurons in 1st layer for each training vector (assumed value)
S2 - number of neurons in 2nd layer for each training vector (assumed value)
W1 - matrix of weight factors for 1st neural layer
W2 - matrix of weight factors for 2nd neural layer .
steam cycle down to those most important and to split the
simulator structure into a greater number of modules.
Difculties in obtaining a satisfactory accuracy
of the networks response
The to- be- solved problem consisted in simulating distri-
bution of parameters of a real object. Functions of the kind
often have a very irregular run, with many discontinuities
resulting from the character of physical phenomena occur-
ring in the power unit, which do not represent only regular
thermodynamical relationships but also many known and
unknown disturbances and small irregularities. If a distri-
bution of a given parameter was correct the network was
able to be trained in generating correct responses. However
in some load intervals when the irregularities were revealed
(e.g. when the set values of the units operational parameters
were very different from those at rated load) the networks
responses appeared loaded by a greater error.
Such situation can be illustrated by the simulation of heat
ow relationships at the rst steam extraction, see Fig.4.
Generally, the steam thermodynamical relationships at this
point of the cycle are very complex and sensitive to many
factors, that has resulted in much greater difculties in ob-
taining a satisfactorily accurate response from the network.
Therefore this point of the cycle has been treated in a special
way : it was taken out from the remaining extractions and
trained separately. The approach made it possible to vacate
some capacity in RAM memory. Also, the computation time
was shortened thus it was possible to increase number of
training iterations. However for some loading states of the
unit the network was not capable of reducing the response
error, satisfactorily. As a result the obtained accuracy of the
network appeared very different, see Fig.5.
Necessity of application of the networks consisted
of many neurons
Complexity of the problem requires the network to be con-
sisted of many neurons. In the cases when the network was
not capable of reducing the error the number of neurons was
increased. The operation usually improved abilities of the
network however it was connected with some drawbacks,
namely :
+ increased loading on the processor computation time was
greater
+ substantially increased loading on RAM memory
+ risk of worsening the networks capability of genera-
lizing.
Fig. 4. Regenerative steam extraction
from the rst stage of HP turbine (the shaded area) .
19
The generalizing capability of the network consists in co-
ping with data sets not included in its training process, which
is of fundamental importance for its usefulness to simulate
phenomena and dependent on the number of neurons, having
its optimum value.
Necessity of extending the training process
The second method of reducing the error is the extending
of training time, apart from the increasing of number of neu-
rons. This is justied only when the reduction goes on. During
training process it often has happened that the network has
reduced the error very slowly though the maximum possible
number of neurons (with respect to RAM memory capacity)
was applied. Then the only way to improve accuracy of the
networks response was to increase number of iterations that
extended training time.
Summing up one can state that in the case in question the
capabilities of the used hardware (a typical personal computer)
were of decisive importance; however they did not substantially
inuence the quality of the elaborated simulator though an ad-
ditional time outlay was necessary.
FEATURES OF THE SIMULATOR
The simulator was composed of 12 neural networks. Each of
them was separately optimized. Due to the hardware limitations
it was necessary to split all the simulated points of the steam
cycle into 3 groups (modules), see Fig.2 and 6.
4 The main steam jet contains the points 1 through 21 located
between the boiler and condenser including the ow part
of the turbines.
4 The rst steam extension is in the point 22 where 3 steam
jets are mixed together : the regenerative steam extension
from the rst stage of the turbine, the steamfromthe sealing
of HP turbine casing and that from the sealing of control
valves.
4 The remaining extensions (points 23 through 30) are loca-
ted in the remaining steam extension pipelines leading to
relevant regeneration heat exchangers.
Fig.6 illustrates the procedure of information ow during
the simulation process. Each of the networks operates separa-
tely and provides only one parameter and only within a given
group of the cycles points. In all the 30 points the distribution
of 4 parameters, i.e. p, m, t, h, can be simulated, or an arbitrary
module and a parameter can be selected to obtain the pressure
parameters from p1 to p21. The user can also select one of the
simulated cycles points to know a concrete parameter value.
CONTROL OF OPERATIONAL
ACCURACY OF THE SIMULATOR
Great attention was paid to continuous control of operation
of the neural networks as they represent rather primitive algori-
thms (not including any interpretation of physical phenomena).
Also, much work was done to reliably present the simulators
accuracy, especially as it appeared to be very different, depen-
ding on a concrete cycles point, simulated parameter and as-

sumed state of loading.
Fig.7 presents distribution of values of the relative error of
mass ow simulation for the main steam jet and all combina-
tions of training data.
During simulation of the parameters an error value is
automatically displayed provided its precise determination is
possible at all. If a set of load parameters does not belong to that
of training data then an expected value of the error is given.
A more precise verication of operational quality of the
simulator was performed by using the DIAGAR analytical
simulator since its responses could be taken as a standard for

Fig. 6. Schematic diagram of the simulators modular structure [6],
where :
h - enthalpy in a given point of cycle [kJ/kg] (resultant)
h1 - h21 - enthalpy put out by 1st module of simulator : main ow
h22 - enthalpy put out by 2nd module of simulator :
1st regenerative steam extraction
h23 - h30 - enthalpy put out by 3rd module of simulator :
remaining regenerative steam extractions
m - mass steam ow rate in a given point of cycle [kg/s] (resultant)
m1 - m21 - mass steam ow rate put out by 1st module of simulator : main ow
m22 - mass steam ow rate put out by 2nd module of simulator :
m23 - m30 - mass steam ow rate put out by 3rd module of simulator :
remaining regenerative steam extraction
N - assumed output of steam power unit [kW]
p - pressure in a given point of cycle [bar] (resultant)
p1 - p21 - pressure put out by 1st module of simulator : main ow
p22 - pressure put out by 2nd module of simulator :
p23 - p30 - pressure put out by 3rd module of simulator :
pk - assumed value of pressure within condenser [bar]
Po - assumed value of fresh steam pressure [bar]
t - temperature in a given point of cycle [C] (resultant)
t1 - t21 - temperature put out by 1st module of simulator : main ow
t22 - temperature put out by 2nd module of simulator :
t23 - t30 - temperature put out by 3rd module of simulator :
To - assumed value of fresh steam temperature [C]
Tp - assumed value of superheated steam temperature [C] .
Fig. 5. Distribution of the relative error of pressure simulation training
at the rst steam extraction, for the whole set of training data.
The mean error : 0.04 %, the maximum error : 0.65% [5] .
Number of input data combinations
R
e
l
a
t
i
v
e

e
r
r
o
r

(
x
1
0
-
3
)
the neural simulator. This way at the expense of some additional
efforts response uncertainty of the network could be reduced
to a minimum. Such operation was performed for the rst
steam extraction and due to its specicity an additional set of
testing data was taken from the DIAGAR and on its basis the
networks generalization accuracy was determined for this
module. In Fig.8 is presented an example of the results of the
above described operation for distribution of mass ow rate
at the rst steamextraction. Atesting data set not contained in
the training data but belonging to the same loading range, was
put into the network.
SUMMARY
+ The neural simulator of the steam power unit was elaborated
to investigate possibility of its application to an on-line
diagnostic system for so complex objects as steam turbine
power plants are.
+ The presented simulator is a tool of the following features :
= simple in use
= fast in operation : determination time of one parameter
in 30 points of the cycle amounts to 30 ms
= sufciently exact
= inexpensive : to its manufacturing only a set of ope-
rational parameters of a considered object, MATLAB
software and a typical personal computer is required.
+ On the basis of the performed task were also analyzed some
practical aspects of the neural modeling method, consti-
tuting its merits and drawbacks related to the considered
application. The most important merit is the possibility
of omitting the long analytical modeling process, hence
lowering the modeling cost; and, the most important draw-
back is that the neural model does not include any physical
interpretation of phenomena.
+ Modelling by means of neural networks is realized on the
basis of existing operational or statistical data. In the case
when character of a phenomenon is complex and multidi-
mensional, and rst of all not quite recognized, as well as
when availability of the data is high, then such conditions
can be taken as very favourable for neural modelling.
In some cases the method is the most suitable, and someti-
mes the only possible. In other cases it can be useful as an
aiding element for analytical models.
+ The tasks assigned to advanced diagnostic systems applica-
ble to large technical systems make that so different appro-
aches as analytical methods, hidden ones (neural networks)
or fuzzy logic, are used for them. Their mutual interaction
is often very favourable for accuracy and credibility of an
obtained result, i.e. assessed technical state of an object.
BIBLIOGRAPHY
1. Guch J. : Verication of selected experimental coefcients of
turbine ow design process , based on standard measurements
of power units (in Polish). Doctorate thesis. Faculty of Ocean
Engineering and Ship Technology, Gdask University of
Technology. Gdask, 1992
2. Gardzielewicz A., Guch J., Uzibo W., Jankowski T.: DIAGAR
software for heat ow diagnostics as a tool to predict repair time
for component devices of turbine power systems. (in Polish).
Proceedings of 5th Domestic Scientic Technical Conference
on Diagnostics of Industrial Processes. agw Lubuski.
September 2001
3. Korbicz J., Obuchowicz A., Uciski D. : Articial neural networks
(in Polish). Academic Publishing House (PLJ). Warszawa, 1994
4. Demuth H., Beale M.: Neural Network Toolbox. 2001
5. lzak-ona J.: Neural simulator of a steam power unit (in
Polish). M.Sc. thesis. Faculty of Ocean Engineering and Ship
Technology, Gdask University of Technology. Gdask, 2004
6. lzak-ona J., Guch J.: Neural simulator of a steam power
unit (in Polish). Proceedings of the Conference on Technical,
Economical and Environmental Aspects of Combined Steam
Systems. Gdask, 2004 .
Fig. 7. Distribution of simulation error of enthalpy parameter
for the main steam jet (21 points of the cycle) [5] .
R
e
l
a
t
i
v
e

e
r
r
o
r

[
-
]
Area of relative error of mass simulation training for
the main steam jet, 20 neurons, 100 iterations .
P
o
in
ts
o
f th
e
c
y
c
le
D
ata com
bination num
ber
Fig. 8. Results of testing the networks generalization capability
for mass ow rate at the rst steam extraction .
Number of intermediate input data combinations
M
a
s
s

[
k
g
]
Generalization testing results for NI97M network .
Justyna lzak-ona, M.Sc.,Eng.
Faculty of Ocean Engineering
and Ship Technology,
Narutowicza 11/12
21
Automation
On 27-30 June 2005 in Warsaw was held 15th Domestic
Conference on Automation, organized by Committee on
Automation and Robotics, and System Research Institute
of Polish Academy of Sciences, and hosted by Warsaw
University of Technology and Industrial Institute of Au-
tomation and Measurements.
The Conferences program contained 196 papers inclu-
ding 6 plenary ones; the rest of them were presented during
16 topical sessions. The session on Technical applications
was the most popular among authors who submitted to it
as many as 40 papers. 13 papers were devoted strictly to
marine engineering problems, namely :
Ship automation systems and CAD methods for them
by Z. Kowalski and M. Drewka (Gdask University
of Technology)
Application of mathematical models in designing
thrusters and electric propulsion systems for ships by
R. Arendt, A. Kopczyski and M.Wojtczak (Gdask
University of Technology)
Structure and selected procedures of a knowledge-
-based system for designing automation of energy
subsystems on ships by R. Arendt, M. Kostrzewski,
Z. Kowalski and E. Van Unden (Gdask University of
Technology)
An algorithm of ship course stabilization based on
a computer model of ship dynamics by P. Borkowski
(Technical University of Szczecin), Z. Zwierzewicz
(Maritime University of Szczecin)
Precise ship control by means of multi-dimensional
though controllers by W. Gierusz (Gdynia Maritime
University)
A system for safe optimum control of ship at sea
by A. ebkowski, R. mierzchalski, M. Tobiasz,
K. Dziedzicki, M. Tomera (Gdynia Maritime Uni-
versity)
Modelling the dangerous and weather areas in ships
routing process by A. ebkowski, R. mierzchalski,
M. Tomera (Gdynia Maritime University)
Simulation of ship precise motion by J. Maecki
(Polish Naval University)
Problems of automation of ships electric energy
systems by R. mierzchalski (Gdynia Maritime
University)
Ship trajectory controller in the aspect of ship steering
in a collision situation by M. Tobiasz, M. Tomera,
A. ebkowski, K. Dziedzicki, R. mierzchalski (Gdy-
nia Maritime University)
Application of back-stepping method to ship motion
control - by A. Witkowska (Gdask University of
Technology), M. Tomera, R. mierzchalski (Gdynia
Maritime University)
Assessment of capability of underwater robot to gene-
rate set-up control outputs by J. Garus (Polish Naval
University)
Modelling dynamic behaviour of underwater robots
by A. ak (Polish Naval University) .
Noise and Vibration
On 8-12 June 2005 at Wigry, a touristic and rest resort
in north-east Poland, was held 7th Conference on :
Active Noise and Vibration Control Methods
It was organized by Faculty of Mechanical Engine-
ering and Robotics, Mining and Metallurgy Academy of
Cracow. The Conferences program contained 56 papers
prepared by the authors representing 19 Polish universities
and scientic research centres, as well as those of Czech
Republic, Finland, France, Lithuania, Slovakia and USA.
The greatest contribution (20 papers) gave scientic wor-
kers of the host university which rst in Poland initiated
research on active vibration and noise control.
Presentation of the papers was carried out
during 8 sessions, namely :
Polish sea-coast scientic circle was
represented by 4 research centres
which presented the following papers :
C Diagnosing the elements of propulsion plant of naval
vessels by means of vibration measurement by A. Char-
chalis, R. Cwilewicz (Gdynia Maritime University)
and A. Grzdziela (Polish Naval University)
C Vibration surveillance in high-speed milling machines
by the spindle speed control by K. Kaliski (Gdask
C Vibration surveillance in the hybrid system by optimum
control of energy performance index - on example
of milling exible details by K. Kaliski (Gdask
C Crack indicators for large power turbo-set monitoring
and diagnostic system by J. Kiciski and S. Banaszek
(The Szewalski Insitute of Fluid-Flow Machinery,
Polish Academy of Sciences, Gdask) .
= Plenary session (4 papers)
= Active vibration control (18 papers)
= Active noise control (7 papers)
= Semi-active control (3 papers)
= Structural control (2 papers)
= Hardware and software for active
noise and vibration control (4 papers)
= Vibration isolation (4 papers)
= Smart materials for adaptive
vibration control (14 papers).
INTRODUCTION
In the paper [1] were presented approximations of ship roll
in regular and irregular waves, obtained on the basis of ship
service parameters and wave parameters. Such approximations
may nd application a.o. in ship voyage routing problems as
well as in controlling ship seakeeping ability during its service.
From the there presented considerations it results that to ap-
proximate ships roll statistical methods and articial neural
networks can be applied, and, the highest approximation ac-
curacy, both within and outside the assumed range of input
parameters, was obtained by applying the functions elaborated
by means of articial neural networks.
In this paper a continuation of the above mentioned research
was presented. It was aimed at checking if articial neural
networks may be used to approximate another ship motion,
namely pitching. Taking into account that in [1] the best results
were obtained for ships roll in irregular waves in this research
the author decided to approximate ship pitching motion in ir-
regular waves only. Additionally, the pitch approximations were
expanded onto the full range of wave encounter angles.
SIGNIFICANT PITCH AMPLITUDES
OF THE SHIP
To approximate the ships pitching the method presented
in [1,6] was applied. The reference data were determined by
using SEAWAY software. In Fig.1 and 2 are presented the
pitch transfer functions calculated by means of the SEAWAY,
in comparison with experimental results; from the comparison
it can be concluded that the SEAWAY provides sufciently ac-
curate results for the kind of ship motions in question.
Approximation of pitching motion of S-175
containership in irregular waves
on the basis of ships service parameters
Tomasz Cepowski
Maritime University of Szczecin
ABSTRACT
This paper presents approximations of pitching motion of S-175 containership in irregular
waves within the full range of ships service parameters. It continues the research described in
[1] dealing with application of articial neural networks to predict ships motions. Reference
pitch values were calculated by means of SEAWAY software based on accurate numerical
methods. Approximating function was elaborated by using the articial neural networks.
Keywords : seakeeping qualities, pitching, articial neural networks,
approximation, ship service parameters, irregular waves
Fig. 1. Pitch amplitudes of the fast cargo ship S.A. van der Stel
of L = 152.5 m , B = 22.80 m , d = 9.14 m , C
B
= 0.563 , Z
G
= 9.14 m ,
sailing in heading waves, acc. [2,3,4,5] .
23
Calculations of the signicant pitch
1/3
were performed,
similarly as in [1], for the example S-175 containership of the
following main dimensions :
length between perpendiculars : L = 175 m
breadth : B = 25.4 m
design draught : d = 9.5 m.
The following ranges of ship service parameters
and wave parameters were taken into account :
ships speed V : from 0 to 20 kn, at every 5 kn
ships draught d : from 5 m to 11 m, at every 2 m
wave encounter angle = 0 (head wave), 10, 20, 30, 60,
90, 120, 150, 160, 170, 180 (following wave)
signicant wave height Hs : from 2 m to 5 m, at every 1 m
characteristic wave period T : from 6.5 s to 14.5 s,
at every 2 s.

As a result of the calculations was obtained the data set
containing 4400 records which was further used to elaborate
pitch approximations. In Fig.3 through 6 are presented selected
Fig. 2. Pitch amplitudes of the fast cargo ship S.A. van der Stel
of L = 152.5 m , B = 22.80 m , d = 9.14 m , C
B
= 0.563 , Z
G
= 9.14 m ,
sailing in following waves, acc. [3,4,5] .
Fig. 3. Signicant pitch amplitudes
1/3
for S-175 containership,
d = var , = var , V = 20 kn , Hs= 5 m , T = 10.5 s .
relationships between the pitch amplitudes calculated this way,
ship service parameters and wave parameters, in different
combinations.
PITCH APPROXIMATION BY MEANS
OF ARTIFICIAL NEURAL NETWORKS
In accordance with [1] the function approximating the
signicant pitch amplitudes
1/3
can be determined by using
the following formula:
(1)
where :
X set of assumed ship service parameters (input)
Y set of values of signicant pitch amplitudes
(output) calculated by using exact methods
1/3
searched function approximating the signicant
pitch amplitudes.
It was assumed that the pitch approximations
had to be elaborated for the following parameters :
1/3
V = var , = var , d = 7 m , Hs = 5 m , T = 10.5 s .
1/3
= var , Hs = var , d = 7 m , V= 20 kn , T = 10.5 s .
1/3
= var , T = var , d = 7 m , V= 20 kn , Hs = 5 m .
Y X
3 1
For determining a function approximating the pitch ampli-
tudes
1/3
the articial neural networks were applied. In this
research the following types of them were tested :
The phase of searching for the most appropriate
network contained the following steps :
To validate and test the networks the set containing 50 %
amount of the variants deleted by sampling from the learning
data set.
The MLP network of the structure : 5 (inputs) x 7 (hidden
neurons) x 1 (output), appeared the most accurate (Fig.7), being
characterized by:
Values of RMS error were calculated
from the formula (2) :
(2)
where :
w
reference values of signicant pitch amplitudes
used for learning or testing a neural network
1/3
approximated signicant pitch amplitudes
n number of records.
The calculations were carried out by using
the package STATISTICA Neural Networks.
The searched function approximating the signicant pitch
amplitudes
1/3
elaborated by means of the above mentioned
neural network was represented analytically with the use of
the equation (3):
(3)
where :
B vector of threshold values :
[2.58 9.69 2.75 - 12.15 - 0.46 - 2.61 - 2.10]
C vector of weighting factor values :
S matrix of coefcients :
P vector of shift values :
[- 0.83333 0 0 - 0.667 - 0.8125]
0
,
1
,
2
coefcients having the values :
0
= - 1.37
= - 0.0065
2
= 0.3268
Sensitivity analysis of the elaborated neural network showed
that all input parameters have been important, at the following
decreasing signicance :
In Fig.8 through 10 is presented the comparison of the pitch
approximations calculated by applying Eq. (3) with the values
calculated by using the exact methods included in the SEAWAY
software. The there presented data were not used for modelling
the neural network. From the diagrams it results that the func-
tion approximating the pitch amplitudes described by Eq. (3)
is very exact and shows the same trends as the data.
+ Multilayer Perceptron (MLP)
of a sigmoidal activation function
+ Generalized Regression Neural Network
(GRNN) a regression network
+ Radial Basic Function Network (RBF).
description of the best network structure by means
of genetic algorithms
learning a network (usually by using the back-pro-
pagation method)
testing a network
assessment of approximation accuracy obtainable
within a network on the basis of the testing data.
( )
n
RMS
w
2
3 1

ship draught d
ship speed V
wave encounter angle
signicant wave height Hs
characteristic wave period T.
d ship draught [m]
Hs signicant wave height [m]
T characteristic wave period [s]
V ship speed [kn]
wave encounter angle [deg]
1/3
signicant pitch amplitude [deg].
2.9538
- 4.5864
- 1.8491
- 4.5384
- 1.3287
3.7736
0.7011
0.01667 0 0 0 0
0 0.0500 0 0 0
0 0 0.0056 0 0
0 0 0 0.3333 0
0 0 0 0 0.1250
Fig. 7. Structure of the articial neural
network approximating pitch amplitudes .
* the wave encounter angle
* the signicant wave height Hs
* the characteristic wave period T
* the ship draught d
* the ship speed V.
0.320 - 0.082 0.111 0.059 0.253 - 0.348 3.170
0.265 - 0.193 1.283 0.182 0.332 0.033 0.291
3.872 16.732 - 2.413 - 20.422 - 0.447 - 3.375 - 1.039
0.390 1.056 0.831 - 0.944 - 1.138 - 0.124 0.133
- 3.097 - 0.575 - 4.895 0.620 0.655 2.671 0.647
A matrix of weighting factors :
+ the smallest learning RMS error = 0.12
+ the smallest testing RMS error = 0.13
2
1 0
) ) T] Hs , , V, [d , (
3 / 1
1
1

|
.
|
\
|

+
= +
+
C
e
B A P S
25
RECAPITULATION
O In this paper are presented the approximations of pitch am-
plitudes, based on the following ship service parameters :
considered within a broad range of the above specied pa-
rameters.
O The pitch amplitude approximating function was elaborated
with the use of articial neural networks and presented
in the analytical form. The function is highly accurate as
compared with the testing data calculated by means of the
exact methods.
O The reference pitch amplitude values were calculated by
using exact numerical methods. On the assumption that the
values do not signicantly differ from real ships pitching
data, the presented approach can be applied to predict ships
pitch on the basis of the data recorded on board the ship.
O The obtained research results conrm that the articial
neural networks can be applied to approximate such ship
response as pitch and roll (presented in [1]) both in the
phase of ships design and operation. [1].
NOMENCLATURE
B ship breadth
C
B
ship block coefcient
d ship draught
F
n
Freude Number
GM initial transverse metacentric height
Hs signicant wave height
L ship length between perpendiculars
RAO Response Amplitude Operator
T characteristic wave period
V ship speed
Z
G
height of ship centre of gravity
wave encounter angle
wave length
1/3
approximated signicant pitch amplitude
w
reference signicant pitch amplitude
BIBLIOGRAPHY
1. Cepowski T. Application of statistical method and articial
neural networks for approximating ships roll in beam waves.
Polish Maritime Research, No. 2, April 2005
2. Gerritsma, J. and Beukelman, W.: Analysis of the Resistance
Increase in Waves of a Fast Cargo-ship. International
Shipbuilding Progress, 18(217), 1972
3. Journe J.M.J. Verication and Validation of Ship Motions
Program SEAWAY. Report1213a, Delft University of
Technology, The Netherlands. 2001
4. Journe, J. M. J.: Motions and Resistance of a Ship in Regular
Following Waves. Technical Report 440, Delft University of
Technology, Ship Hydromechanics Laboratory, The Netherlands.
(internet: http://dutw189.wbmt.tudelft.nl/~johan or http://www.
shipmotions.nl). 1976
5. Journe, J. M. J.: Motions, Resistance and Propulsion of a Ship
in Longitudinal Regular Waves. Technical Report 428, Delft
University of Technology, Ship Hydromechanics Laboratory,
The Netherlands. (internet: http://dutw189.wbmt.tudelft.
nl/~johan or http://www.shipmotions.nl). 1976
6. Staszewska K., Cepowski T.: Prediction of ships roll by means
of articial neural networks (in Polish), Scientic bulletins of
Szczecin Maritime University (Zeszyty Naukowe AM) No. 74.
Szczecin, 2004
7. Szelangiewicz T., Cepowski T.: Application of articial neural
networks to investigation of ship seakeeping ability, Part 1,
Polish Maritime Research, Vol 8, No. 3, September 2001 .
Fig. 8. The approximations of signicant pitch amplitudes, compared with
the testing values, for : = var , V = 20 kn , T = 12.5 s , Hs = 5 m , d = 11 m .
Fig. 9. The approximations of signicant pitch amplitude, compared with
the testing values, for : V = var , = 0 deg , T = 12.5 s , Hs = 5 m , d = 11 m .
Fig.10. The approximations of signicant pitch amplitudes, compared with
the testing values, for : d = var , = 0 deg , T = 12.5 s , Hs = 5 m , V = 20 kn .
ship speed
ship draught
wave encounter angle,
and the wave parameters :
signicant wave height
characteristic wave period,
Tomasz Cepowski, D.Sc.,Eng.
Institute of Marine Navigation,
Way Chrobrego 1/2
70-500 Szczecin, POLAND
INTRODUCTION
Ensuring adequate technical parameters of port water-
ways as well as the development and modernisation of the
port infrastructure are the main issues related to the safety
of maritime transport. Although fairway parameters such as
depths and widths, different for each port, affect its safety and
efciency, management of waterways is equally important.
Safe and effective port waterways management depends on
trafc density information and actual meteorological and hy-
drological conditions.
Improvements in the safety of ports and waterways are ne-
cessary for several reasons including demands of growing inter-
national trade, the trend toward larger ships and faster loading
and unloading operations, the presence of oil and hazardous
cargo in congested and heavily populated areas and concerns
about maritime accidents that may cause environmental dama-
ge. These improvements can be realised through investments
both into traditional aids of navigation (beacons, buoys, lights)
and advanced maritime information systems like real-time in-
formation on weather conditions, pilotage navigation systems
and decision support systems for a VTS Centre.
Although the term Vessel Trafc Service was internatio-
nally used for the rst time in the 1970s, systems supporting
vessel trafc control were established much earlier [11].

First
local initiatives were taken just after the Second World War.
Liverpool was the rst port in the world to receive (in 1948)
a monitoring station and radio equipment. Over the next fty
years such services were introduced by port authorities and
coastal state administrations and at present more than 250 such
systems ensure regular control in port areas, port approaches
and narrow trafc lanes [3].
The VTS idea is based on establishing services for ships
trafc safety and efciency improvement. The scope of tasks
may vary from simple information exchange in ship-land com-
munications to trafc management, i.e. planning such vessels
routes and passages which minimise risk of collision. The as-
sessment of collision risk is based on the information on actual
meteorological and hydrological conditions, trafc density,
available aids of navigation and technical infrastructure [5].
The creation of VTS systems made it is necessary to dene
rights, obligations and responsibilities of shore-based centres,
vessel masters and pilots. A VTS, with its specialised knowled-
ge of a given waterway is responsible for trafc management in
the area. On the other hand, the master, with his knowledge of
vessel behaviour and professional skills, is responsible for the
safety of the vessel. With taking into account the different- but
related - responsibilities, all instructions from the shore centre
to the vessel must be result-oriented, while the details of their
execution fall on the master and pilot.
KNOWLEDGE ACQUISITION
AND REPRESENTATION
Knowledge acquisition can be dened as a formalised,
heuristic process of gathering information, data or explanations
concerning the domain of relevant knowledge [2]. Experts are
the main source of such knowledge. They cannot be usually
omitted in the knowledge acquisition process.
The effectiveness of expert system is determined by the
process of knowledge acquisition. In this process, the arti-
culation of knowledge is a topic of great importance. In the
Assessment of the safety level
of a ships passing manoeuvre in the fairway
Marek Narkiewicz
ABSTRACT
Safe two-way vessel trafc may have some limitations. Fairway width, vessel dimensions,
weather conditions, trafc density and other factors must be considered crucial and con-
straining safe passage of the vessel. The possible sea trafc regulation made by a Vessel
Trafc Service (VTS) shore centre, is often based on criteria included in harbour regulations.
This may not always be an optimum solution. This paper presents a concept and elements
of a method for assessment of safe two-way trafc, based on experts (pilots) knowledge
and experience. The goal is to support the decision making process in a VTS centre. For
this purpose tools of knowledge acquisition and representation as well as tools of statistical inference are
used. The method is based on the results of the expert research which was performed with the participation
of pilots who handle vessels on the winoujcie-Szczecin fairway.
Keywords : safety of navigation, Vessels Trafc Service (VTS)
27
presented research the method of knowledge acquisition based
on examples was used. This is a method of inductive nature, in
which a general model is constructed by using induction rules
and is based on a set of examples [9].
Experts questionnaires are one of the most popular heu-
ristic methods of knowledge acquisition. The independence
of experts opinion and the anonymity of judgements made
by experts working in isolation are distinctive features of this
method. The anonymity ensured during the research allows an
expert to express an opinion which is not in conformity with
the most common one.
The terms expert and expert opinion do not have a uni-
form interpretation. For instance, an expert can be considered
as a person invited to take part in the research because of his or
her knowledge, personality, professional experience etc [4].
An expert can be dened as a very skilful person who has
had much training and has knowledge in a specic eld. This
expert is the provider of an opinion in the process of experts
opinion elicitation. An expert opinion can be dened as a for-
mal judgement of an expert on a matter in which his or her
advice is sought. Such opinion is a subjective assessment,
evaluation, impression or estimation of the quality or quantity
of something of interest that seems true, valid or probable to
the experts own mind [2].
The research subject described in this paper is the part of
specic experts knowledge learned in the training process.
It consists of all the particular heuristics and shortcuts that
a trained professional has learned to use in order to perform
better. His or her practice is anchored in the theory which is
also needed but could be only the base [6].
The term knowledge representation could be used to
dene the formalised process of transferring, writing and ga-
thering the knowledge meant as a description of surrounding
world, consisting of facts and relations. Such knowledge re-
presentation must provide a proper structure of description and
interpretation. The process has to be formalised when experts
knowledge is to be used in a decision support system.
There is no one, universal, widely accepted way of know-
ledge representation. Methods of representation must take into
consideration a variety of knowledge elements. However future
application is also very important. There are only some features
that can be required, namely : transparency, accuracy, efcacy,
neutrality and adequacy [12].
Some authors use the term uncertain knowledge repre-
sentation for the denition of methods of inference used in
decision support systems. Those authors dene uncertainty
as a lack of information needed for decision making. Sources
of uncertainty are : inaccuracy of measuring methods, the
existence of parameters omitted as well as the lack of proper
experts knowledge [7].
To dene means of interpretation of accepted uncertainty
measures is a difcult and important task. The use of probability
calculus is one admissible solution. Probability is commonly
accepted as a quantitative measure of uncertainty both for pre-
mises and for rules of inference.
Statistical methods of analysis (descriptive and inference) are
a very useful tool in the perception of structure, interdependence
and dynamics of a phenomenon [8]. These methods range from
simple explanatory data analysis and signicance testing to hi-
ghly complicated mathematical modelling techniques. Some of
them help to summarise efciently large amount of data while the
others help to understand the effects of a number of variables on
another variable of interest. One should always remember that
statistical inference belongs, besides inference by analogy, to
the group of probabilistic inferences, which means that on the
basis of true premises we can come to false conclusions [13].
PRINCIPLES OF THE METHOD
The presented method was based
on the following principles :
+ Safety level is the quantitative feature of each manoeuvre.
It depends on many factors but mainly on: manoeuvring
area parameters, vessels parameters, executed manoeuvre
parameters, and those of actual hydrological and meteoro-
logical conditions.
+ Safety level index is the quantitative measure of the safety
level.
+ Safety level can be rated by an expert (pilot) because of his
or her knowledge and professional experience. The experts
knowledge can be gained by using experts questionnaires.
+ The questionnaire used in this research was built on the
following principles :
+ The questionnaire contains an experts opinion which
is a subjective assessment (evaluation) that seems true
and valid to the experts own mind.
+ The experts opinion expressed by the gure ranging
from 1 (safe manoeuvre) to 10 (unsafe manoeuvre),
is the quantitative measure of the safety level of ships
passing manoeuvre.
+ Evaluation of the safety level made by experts is based on
specic experts knowledge learned in the training process
in order to perform better.
INVESTIGATIONS
The expert research (by using questionnaires) was per-
formed with the participation of pilots handling vessels on
the winoujcie-Szczecin fairway. It aimed at answering two
main questions:
C Is the safety level of manoeuvres of two vessels passing each
other affected only by their main dimensions or by actual
hydrological and meteorological conditions as well ?
C Is the safety level of manoeuvres of two vessels passing each
other affected also by the section of the fairway in which
the manoeuvre is executed? (i.e. are there sections on the
fairway where manoeuvre execution is safer not only due
to the canal width).
The safety level was based on pilots assessment with the
use of 10-degree scale. Only the vessels frequently calling at
port of Szczecin were taken into consideration.
One of the main tasks of the research was to determine
the inuence of vessels dimensions and hydrological and
meteorological conditions on safety level (pilots asses-
sment). Nine following factors (independent variables) were
assumed :
* length, width and draught of the vessel entering the port
* length, width and draught of the vessel leaving the port
* visibility
* wind force
* time of day (daylight or night).
The factor not taken into consideration was a kind (dan-
gerous / non-dangerous) of the cargo on board. The results of
previously performed studies show a very low priority of this
factor for pilots. This factor must of course be implemented
in harbour regulations but its inuence on pilots assessment
cannot be considered signicant. Pilots opinion in this mat-
ter may also be affected by his or her professional experience
(e.g. gained in handling oil and/or chemical tankers) [10].
The research task was to nd out if the inuence of all above
mentioned factors is equal in all fairway sections.
According to the Regulation of Ministry of Infrastructure
[14] winoujcie-Szczecin waterway should have the width of
90 m with the depth of 10.5 m (with the exception of a part of
the port of winoujcie and one branch of the waterway near
the port of Police where the required width is 150 m). It is not,
however, a uniform waterway. It consists of canals, dredged
stretches of rivers (Odra, wina) and the Gulf of Szczecin. For
the purpose of this research the waterway was divided into 11
sections shown in Fig.1.
Each questionnaire consists of assessment of several ma-
noeuvres. Thirty two pilots took part in the research. Nine the
same manoeuvres were assessed by all the pilots.
When establishing the range of vessel dimensions two
restrictions were considered. On the one hand the assessed
manoeuvres should be those permitted by harbour regula-
tions (slight violations can be accepted). Firstly this restric-
tion is due to the fact that manoeuvres not permitted by the
regulations could be considered by pilots as unsafe, secondly
because only such manoeuvres can be considered as per-
formed by pilots in the past. On the other hand the results of
trial questionnaire showed that the assessments of manoeu-
vres of small vessels (about 80 m in length) never exceeded
2 in 10-degree scale regardless of the waterway section. This
may lead to the conclusion that for such vessels trafc control
is not necessary.

THE RESULTS
Assessing the safety level of a passing manoeuvre the pilot
must take into consideration a number of factors. To nd out
how pilots assessment is performed, one must answer several
questions regarding the signicance and inuence of particu-
lar factors. What is the effect of vessels dimensions ? Which
dimensions can be considered as inuential ? Could the same
manoeuvre executed in different meteorological conditions be
assessed at a different safety level ? Is the visibility restricted
to 5 Nm an inuential factor ? How does the night time affect
pilots opinion ? To answer questions like these the regression
analysis, the most widely applied of all statistical techniques,
can be used.
The regression model aims to determine how a set of varia-
bles (in this research : vessels dimensions, hydrological and
meteorological conditions) is related to another variable
in question (safety level). Like other models it represents
a simplied description of reality, i.e. that neglecting elements
non-essential for the researcher. Such simplication makes the
problem easier to understand, but one should bear in mind that
an excessive simplication needs unrealistic presumptions.
An equation or equations of the model are usually rep-
resented in a standardised form containing the looked-for
(dependent) variable and inuencing (independent) variables
(predictors).
Y = b
0
+ b
1
X
1
+ b
2
X
2
+ ... + b
n
X
n
(1)
where :
Y dependent variable
X
1
, X
2
... X
n
independent variables
b
0
, b
1
... b
n
equation coefcients.
To select the best set of predictors is an important problem.
Such variables should take into consideration the aim of the
research descriptive or prognostic one, and assure a sensible
interpretation and description of the dependent variable with
sufcient accuracy.
The approach chosen to solve the problem of the variable
selection is sometimes called the experiment planning or
backward elimination. Such an approach is convenient when
the best-tting regression equation is not needed, but only to
nd out the signicance of independent variables in predicting
the dependent variable (pilots assessment) is needed. The
vessels dimensions and parameters describing hydrological
and meteorological conditions were assumed the independent
variables. Almost all variables are of qualitative nature. The
only categorical (dummy) variable is a part of the day : day or
night. To identify its signicance it is denoted as follows :

0 day
1 night.
According to the taken presumptions the description is
important only when it conrms statistical relations between
the predictors and pilots assessment. Such relation can be
conrmed by using the t-Student test used to verify a hypothe-
sis concerning regression equation coefcients. The assumed
signicance level was 0.05.
H
0
: b
i
= 0 and the alternative hypothesis H
1
: bi 0
The optimum decision allows to reject H
0
. This conrms
the signicance of the variable. The acceptance of H
0
means
that the dependent variable is not affected by this predictor or
that the inuence of this predictor cannot be precisely deter-
mined. The conrmed signicance of the predictors allows to
incorporate them into the equation. In the regression model
Fig. 1. winoujcie-Szczecin fairway and its sections .
No of
section
Part of the fairway
1 Entrance
2
Harbour master ofce-northern
promontory of Mielin island
3
Northern promontory of Mielin
Island - Karsibrz
4 Karsibrz - fairway gate No I
5 Fairway gate No I - buoys 7-8
6 Buoys 7-8 - Makw
7 Makw - dolphin Krpa Dolna
8
Dolphin Krpa Dolna - dolphin
Radu Dolna
9 Dolphin Radu Dolna - Inoujcie
10 Inoujcie - Orli Przesmyk
11 Orli Przesmyk - Basen Grniczy
29
the correlation between independent and dependent variables
is important, but it is also important to select such variables
which are not mutually correlated.
Predictors which describe vessels should take into conside-
ration dimensions of both vessels because one can not assume
safety level of passing manoeuvre of vessels knowing length,
breadth and draught of one vessel only. However it is possible
to include into the equation the variables which are sums of
corresponding dimensions, but rst it must check if there is no
correlation between the variables.
When verifying if a correlation between independent varia-
bles does not exist the correlation factors matrix shown in
Tab.1 can be used. These factors are measures of correlation
between pairs of independent variables. If two variables are
strongly correlated it can affect estimation procedure of re-
gression parameters.
The results presented in Tab.1 show strong correlation
between two variables: sum_of_lengths and sum_of_breadths.
This result makes it impossible to include into the equation
the variables which are sums of corresponding dimensions
(length and breadth). They can be replaced with a new predic-
tor, i.e. sum of the products of the two dimensions, called the
sum_of_waterplanes.
sum_of_waterplanes = L
1
B
1
+ L
2
B
2
(2)
where :
B
1
, B
2
breadths of considered vessels
L
1
, L
2
lengths of considered vessels .
Tab. 1. Correlation of pairs of variables .
VARIABLE
Sum_of_
lengths
Sum_of_
breadths
Sum_of_
draughts
Night Visibility
Wind
force
Sum_of_
lengths
1.0000 0.9124 0.7222 - 0.0000 0.0000 0.0000
Sum of_
breadths
0.9124 1.0000 0.7394 0.0000 0.0000 - 0.0000
Sum of_
draughts
0.7222 0.7394 1.0000 - 0.0000 0.0000 - 0.0000
Night - 0.0000 0.0000 - 0.0000 1.0000 0.0542 - 0.1259
Visibility 0.0000 0.0000 0.0000 0.0542 1.0000 - 0.3937
Wind
force
0.0000 - 0.0000 - 0.0000 - 0.1259 - 0.3937 1.0000
Acceptance of the ve variables: sum_of_waterplanes,
sum_of_draughts, night, visibility, and wind force allows to
create the following testing hypotheses :
H
0
1

: b
sum_of_waterplanes
= 0
and alternative hypothesis
H
1
1
: b
sum_of_waterplanes
0
H
0
2
: b
sum_of_draughts
= 0
H
1
2
: b
sum_of_draughts
0
H
0
3
: b
night
= 0
H
1
3
: b

night
0
H
0
4
: b
visibility
= 0
H
1
4
: b
visibility
0
H
0
5
: b
wind force
= 0
H
1
5
: b
wind force
0
In Tab. 2 are shown the results of the tests of signicance of
the predictors for the section no. 4 and 6 of the waterway. The
results of the tests for the other sections are similar.
Tab. 2. Results of tests of signicance of predictors
for the section no. 4 and 6 .
Section no. 4
PREDICTOR b t
(204)
p
Sum_of_waterplanes 0.000223 2.20254 0.028747
Sum_of_draughts 0.346985 3.45112 0.000678
Night 0.096538 0.35815 0.720601
Visibility - 0.113797 - 3.06849 0.002444
Wind force 0.008238 0.40967 0.682481
Section no. 6
PREDICTOR b t
(204)
p
Sum_of_waterplanes 0.000266 3.08243 0.002337
Sum_of_draughts 0.416491 4.85477 0.000002
Night 0.001524 0.00662 0.994721
Visibility - 0.066313 - 2.09560 0.037351
Wind force 0.021087 1.22893 0.220513
The presented results conrm that the three predictors:
sum_of_waterplanes, sum_of_draughts and visibility are signi-
cant. H
0
hypotheses for the other variables cannot be rejected.
This may lead to the conclusion that wind force (because of
a short time of manoeuvre) and the time of day does not affect
pilots assessment.
An obvious way to assess the regression equation is to
assess how well a set of independent variables correlates with
the dependent variable. Measures of tting are the variance
and standard deviation of the rests, and the coefcient of de-
termination. Measures of accuracy are standard deviations of
the coefcients.
The coefcient of determination r
2
is the percentage of
total variance explained by the model. It is an indicator of the
model t [1] :
(3)
where :
SSE sum of squares of errors
SST total sum of squares .
This coefcient can be also adjusted with the number
of degrees of freedom; it has the following form :
(4)
where :
n number of observations
k number of predictors .
Another measure of model tting
is the root mean square error (or mean standard error) :
(5)
where :
MSE the mean square error.
(6)
where :
y
j
actual value of dependent variable

j
predicted value of dependent variable .
SST
SSE
r
2
=
) r 1 (
1 k n
1 n
1 r
2 2
) adj (

=
MSE s =
) 1 k ( n
) y y (
) 1 k ( n
SSE
MSE
n
1 j
j
^
j
+

=
Tab. 3. Coefcient of determination, adjusted coefcient of determination,
F-statistic and mean standard error, for the equation of three variables :
sum_of_waterplanes, sum_of_draughts and visibility .
Number
of section
r
2
r
2
(adj)
F
(3.206)
p s
1 0.34344 0.33387 35.91831 < 0.00001 1.81659
2 0.25954 0.24876 24.06834 < 0.00001 1.64267
3 0.22717 0.21591 20.18418 < 0.00001 1.89743
4 0.23437 0.22322 21.02033 < 0.00001 1.90984
5 0.29764 0.28741 29.09912 < 0.00001 1.84703
6 0.34136 0.33177 35.58822 < 0.00001 1.63436
7 0.32367 0.31382 32.86169 < 0.00001 1.69795
8 0.31178 0.30176 31.10757 < 0.00001 1.63894
9 0.32447 0.31463 32.98172 < 0.00001 1.56115
10 0.31313 0.30312 31.30318 < 0.00001 1.60150
11 0.40054 0.39181 45.88091 < 0.00001 1.51395
Tab.3 shows values of the coefcient of determination, the
adjusted coefcient of determination and the mean standard
error for the equation of three variables : sum_of_waterplanes,
sum_of_draughts and visibility. It also shows an F-statistic
analysis of variance, which conrms the relation between
the dependent variable and any of independent variables.
The equation is based on 210 examples assessed by different
pilots.
The obtained values of the coefcient of determination
show that only a small percentage of total variance can be
explained by the model. Nevertheless, obtaining the best-tting
prediction was not the purpose of this part of research; it was
the conrmation if the predictors affect the dependent varia-
ble. Due to such factors like the inaccuracy of measurement
(experts assessment based on specic professional experience),
existence of neglected parameters of the phenomenon in qu-
estion (i.e. water current, ships speed) or the lack of a proper
experts knowledge, the used model is not able to account for
everything, thus errors are unavoidable.
The equation ts the data better when the set of three varia-
bles the signicance of which was conrmed by tests, is inclu-
ded in the equation and mean values of all pilots assessments
are taken into account. Tab.4 shows values of the coefcient
of determination, the adjusted coefcient of determination, the
F-statistic analysis of variance and the mean standard error for
the equation with the three variables. The equation is based on
nine examples, each assessed by all 32 pilots. The mean values
of assessments are taken into account.
Tab. 4. Coefcient of determination, adjusted coefcient of determination,
F-statistic, and mean standard error, for the equation with three variables :
sum_of_waterplanes, sum_of_draughts and visibility. Mean values of all
pilots assessments are taken into account .
Number
of section
r
2
r
2
(adj)
F
(3,5)
p s
1 0.92816 0.88506 21.53373 0.00274 0.52673
2 0.96134 0.93814 41.44007 0.00059 0.28332
3 0.94253 0.90804 27.33268 0.00158 0.43437
4 0.91715 0.86744 18.45003 0.00390 0.46360
5 0.88738 0.81981 13.13251 0.00831 0.61021
6 0.95045 0.92072 31.96948 0.00109 0.48449
7 0.92279 0.87646 19.91910 0.00328 0.54660
8 0.92991 0.88785 22.11112 0.00258 0.53383
9 0.93428 0.89485 23.69487 0.00220 0.48674
10 0.92943 0.88709 21.95142 0.00263 0.52384
11 0.92411 0.87858 20.29546 0.00314 0.57119
Tab.5 shows the constant and coefcients b attributed to the
variables: sum_of_waterplanes, sum_of_draughts and visibility
for all eleven sections of the winoujcie-Szczecin fairway. Fig.2
shows the coefcients b attributed to the same three variables
for all eleven sections of the winoujcie-Szczecin fairway.
Tab. 5. The coefcients b attributed to the variables: sum_of_waterplanes
(b
1

), sum_of_draughts (b
2

) and visibility (b
3

) for all eleven sections of the
winoujcie-Szczecin fairway (b
0
- constant). Mean values of all pilots
assessments are taken into account .
Number of
section
b
0
b
1
b
2
b
3
1 2.253766 0.000615 0.115413 - 0.169931
2 2.117815 0.000429 0.056652 - 0.157979
3 0.564922 0.000479 0.229381 - 0.139367
4 0.369852 0.000415 0.196733 - 0.128929
5 - 1.287264 0.000308 0.342393 - 0.154076
6 - 0.067514 0.000548 0.322482 - 0.152738
7 - 1.034664 0.000375 0.381633 - 0.142633
8 - 0.213562 0.000371 0.414399 - 0.137965
9 0.243243 0.000429 0.274761 - 0.168543
10 0.489656 0.000457 0.295718 - 0.157638
11 - 0.206354 0.000388 0.452058 - 0.099156
Fig. 2. The coefcients b attributed to the variables : sum_of_waterplanes and sum_of_draughts .
Section of the fairway
0.00000
0.00010
0.00020
0.00030
0.00040
0.00050
0.00060
0.00070
1
Entrance
2
Harbour
master
office -
- northern
promontory
of Mielin
island
3
Northern
promontory
of Mielin
island -
- Karsibrz
4
Karsibrz -
- Fairway
5
Fairway
- buoys 7-8
6
Buoys 7-8 -
- Makw
7
Makw -
- dolphin
Krpa Dolna
8
Dolphin
Krpa
Dolna -
- dolphin
Radu Dolna
9
Dolphin
Radu Dolna -
- Inoujcie
10
Inoujcie -
- Orli
Przesmyk
11
Orli
Przesmyk -
- Basen
Grniczy
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
sum_of_waterplanes
sum_of_draughts
C
o
e
f
f
i
c
i
e
n
t
b
f
o
r
t
h
e
p
r
e
d
i
c
t
o
r
t
o
1
s
u
m
_
o
f
_
w
a
t
e
r
p
l
a
n
e
s
C
o
e
f
f
i
c
i
e
n
t
b
f
o
r
t
h
e
p
r
e
d
i
c
t
o
r
t
o
2
s
u
m
_
o
f
_
d
r
a
u
g
h
t
s
gate No I
gate No I -
31
The values of coefcients for different sections are similar
except for the sections no.2 and 5. In the section no.2 which
includes the fairway of the port of winoujcie with the depth
of 14.3m, the value of the coefcient b attributed to the variable
sum_of_draughts is distinctly lower, compared to those related
to the other sections. In the section no.5 which leads through
the Gulf of Szczecin, the value of the coefcient b attributed
to the variable sum_of_waterplanes is distinctly lower than
for the other sections, while the coefcient b attributed to the
variable sum_of_draughts can be regarded as affecting the
pilots opinions.
Example :
with three predictors assumed, the equation
for the section no.7 is as follows :
(assessment) = -1.03466 + 0.00037 (sum_of_waterplanes) +
+ 0.38163 (sum_of_draughts) - 0.14263 (visibility)
The manoeuvre of two ships passing each other and having
the parameters : L
1
=150 m, B
1
=18 m, T
1
=7 m; L
2
=120 m,
B
2
=15 m and T
2
=5 m, respectively (shown in Fig. 3 with solid
line), in the visibility of 10 Nm, was assessed by the pilots at
the value of 3.81 in 10 - degree scale. The same manoeuvre in
the visibility reduced to 0.5 Nm was assessed at the value of
5.16. Also, can be compared two manoeuvres given with the
same mark in two different visibility conditions, i.e. the good
visibility (10 Nm) and that reduced to 0.5 Nm. The assessment
value of 3.81 calculated for the above described manoeuvre
can be assigned to the manoeuvre in the reduced visibility only
when the dimensions of ships are adequately smaller. With the
draught unchanged, the length and beam of one of the vessels
should be equal to L
2
= 88 m and B
2
= 10 m (shown in Fig.3
with the dotted line).
CONCLUSIONS
O Assessing a given situation, when manoeuvring in restricted
areas, the pilot must take into consideration a great number
of factors. The creation of a decision support systems based
on experts knowledge and experience, seems to be a use-
ful and helpful solution. Such systems can solve complex
problems of a very specic nature with results comparable
to those provided by an expert, for example a pilot.
O When a trafc regulation is established in a specic area,
decisions related to planning the passage of vessels are not
made by the pilot but by a person who has no such profes-
sional shiphandling experience. His or her decisions are
based on harbour regulations or internal procedures which
must be a simplication of pilots experience.
O The knowledge acquisition process is laborious and long-
-lasting, so the creation of decision support system is
reasonable only when it has to be used by a great number
of users and for a long time.
O There are opinions that the main task of an expert is to
tackle identication of a hazard or risk. Anybody with
great professional experience can easily point out situations
connected with hazard or risk. The results presented in the
paper show that an expert opinion can be also used for risk
classication.
O Finally, all activities undertaken by all the persons engaged
in safety matters in restricted waters can be treated as the
activities undertaken within one system which needs inter-
dependent decision - making and operations. The system
in which decisions made by persons in charge of the vessel
and a shore-based VTS centre interact each other, really
exists. The interaction must be taken into consideration by
all parties because decisions made by one party affect the
other parties in an intended or unintended manner.
NOMENCLATURE
b equation coefcient
B vessels breadth
F F-statistic
L vessels length
p signicance level
r
2
coefcient of determination
r
2
(adj)
adjusted coefcient of determination
s mean standard error
t t-statistic
T vessels draught
VTS Vessels Trafc Service
Y dependent variable
X independent variable.
BIBLIOGRAPHY
1. Aczel A. D.: Statistics in management (in Polish), Scientic
Publishers (Wydawnictwo Naukowe) PWN. Warszawa, 2000
2. Ayyub B. M.: Elicitation of Expert Opinion for Uncertainty and
Risk, CRC Press LLC. Boca Raton, 2001
3. Boisson P.: Safety at sea Policies, Regulations & International
Law, Bureau Veritas. Paris, 1999
4. Cielak M.: Predicting in economy. Methods and applications
(in Polish), Scientic Publishers (Wydawnictwo Naukowe)
PWN. Warszawa, 2001
5. Filipowicz W.: Problems of ship trafc management PLA
(Population Learning Algorithms) approach (in Polish), Reports
of Navigation Department, Gdynia Maritime University
(Prace Wydziau Nawigacyjnego Akademii Morskiej) No. 13.
Publishing House of Gdynia Maritime University. Gdynia, 2002
6. Klein M. R., Methlie L. B.: Knowledge-based Decision Support
Systems with Application in Business. John Wiley & Sons Ltd.
Chichester, 1995
7. Korbicz J., Kocielny J. M., Kowalczuk Z., Cholewa J.:
Process diagnostics. Models. Articial intelligence methods.
Applications (in Polish). Scientic Technical Publishers
(Wydawnictwo Naukowo-Techniczne). Warszawa, 2002
8. Luszniewicz A., Saby T.: Statistics. Theory and applications
(in Polish). 2nd ed., C.H. Beck Publishers. Warszawa, 2003
9. Mulawka J.: Expert systems (in Polish). Scientic Technical
Publishers (Wydawnictwo Naukowo-Techniczne). Warszawa,
1996
9
0
m
Fig. 3. Passing manoeuvre .
10. Pietrzykowski Z., Narkiewicz M.: Implementation of marine
regulations in decision support system for vessel trafc services.
Maritime Engineering & Ports. WIT Press. Southampton, 2002
11. Wawruch R., niegocki H.: System of ship trafc surveillance
as a body for control and organization of ship trafc at sea
(in Polish). Reports of Navigation Department, Gdynia Maritime
Academy (Prace Wydziau Nawigacyjnego Wyszej Szkoy
Morskiej w Gdyni) No. 4. Publishing House of Gdynia Maritime
Academy. Gdynia, 1997
12. Zieliski J. S.: Intelligent systems in management. Theory
and practice (in Polish). Scientic Publishers (Wydawnictwo
Naukowe) PWN. Warszawa, 2000
13. Ziembiski Z.: Practical logics (in Polish). Scientic Publishers
(Wydawnictwo Naukowe) PWN. Warszawa, 1994
14. Polish Ministry of Infrastructure. Ofcial Journal No 4, item 41.
Warszawa, 2003 .

Marek Narkiewicz, M.Sc.,Eng.
Department of Marine Trafc Engineering,
Institute of Marine Trafc Engineering,
Way Chrobrego 1/2
70-500 Szczecin, POLAND
On 16-19 May 2005 at Puck, a Polish town
on the coast of the Gulf of Gdask,
was held 5th in Poland
European Acoustics Association (EAA)
Symposium on Hydroacoustics.
It was organized by Polish Naval University, Gdynia.
As usually, the main topics of the scientic meeting were
the following :
34 submitted papers were devoted to research results
and theoretical considerations presented by scientic wor-
kers from 11 Polish scientic research centres including
Gdask University of Technology (10 papers), and 8 fore-
ign centres including 4 Russian ones (4 papers) and those
of Austria, Canada, Turkey and United Kingdom (1 paper
each). All the papers have been very carefully published
(in Polish) as the Symposiums proceedings by Polish
Acoustical Society, Gdask Division, in the Annual Jour-
nal Hydroacoustics, vol.8, Gdynia 2005.
SHA 2005
* Acoustics in sheries
* Acoustics in marine environment
* Non-linear acoustics
* Sound propagation in the sea modelling
* Radiated and ambient noise
* Sonar systems
* Signal and data processing
* Transducers and instrumentation.
The next successive
Domestic Scientic Conference on :
Scientic and technical problems
of professional sailing
was organized by the Faculty of Motor Cars and Heavy
Machinery, Faculty of Mechanics, Energy and Aeronautics
of Warsaw University of Technology together with the
Faculty of Mechanical Engineering of Gdynia Maritime
University.
It was held on 2-9 April 2005 on board the sailing
yacht Pogoria during its voyage on the route : Naples
Castellamare Civittaveccia Livorno Geneva, and
its programme contained the presentation of 25 papers pre-
pared both by scientic workers and students. The papers
were divided into four topical groups :
+ Calculation methods, measurements and materials
10 papers including 3 prepared by students :
Measurement of hull form by using tachimetric
methods by A. Klawikowska and B. Puchowski
under supervision of J. Kozak, D.Sc. (Gdask Uni-
versity of Technology)
Numerical analysis of a keel-rudder system by using
Fluent software by A. Sentkowska and J. Broni-
szewski (Warsaw University of Technology)
Analysis of inuence of spreader inclination angle on
magnitude of compressive force acting on the mast
and stress distribution within the mast by M. Za-
godon (Warsaw University of Technology) .
+ Construction, building and general arrangement
6 papers including 4 prepared by students :
A fast water craft driven by human muscles by
W. Leniewski and K. Niklas (Gdask University
of Technology)
Design study of a small yacht by Z. Madej (Sile-
sian University of Technology)
Tests of a prototype sailing yacht by D. Marku-
szewski and M. Wdoowski (Warsaw University
of Technology)
Flybridge as an element of Sunreef 60 turistic
catamaran by E. Perzyk (Warsaw University of
Technology) .
The greatest number of papers (8) was prepared by
representatives of Warsaw University of Technology,
4 papers by authors from Mining- Metalurgical Academy,
and 2 papers by authors from Gdynia Maritime University.
Authors from Gdask University of Technology, Koszalin
Technical University, Silesian University of Technology
and Technical-Humanistic Academy of Bielsko-Biaa pre-
sented one paper each.
A sailing conference
+ Safety and operation 6 papers
+ Mechanical drives 3 papers

Polish Maritime Research

Uploaded by

Copyright:

Available Formats

Polish Maritime Research

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Polish Maritime Research

Uploaded by

Copyright:

Available Formats

Address of Publisher

& Editor's Ofce :

source of the quantity .

are classied with accounting

in the successive time-step

. As a result of the application of one of the

12 POLISH MARITIME RESEARCH, No 1/2006

- vector of integrand functions of a set of ordinary

- vector of function values in nodes of calculation grid

You might also like