You are on page 1of 11

Collision and Grounding of Ships and Offshore Structures – Amdahl, Ehlers & Leira (Eds)

© 2013 Taylor & Francis Group, London, ISBN 978-1-138-00059-9

An improvement on a method for estimating number of collision


candidates between ships

F. Kaneko
National Maritime Research Institute, Mitaka, Tokyo, Japan

ABSTRACT: Methods for estimation of frequency of collision candidates have been developed by several
researchers, such as Fujii, Pedersen and Kaneko etc. Those methods are thought to be able to deal with wide
ranging situations of collision candidates. However the case of small crossing angle has not been dealt with to a
satisfactory extent. This paper introduces a method to estimate the number of collision candidates in a crossing
between twowater ways which cross with a small angle. Prior to this a holistic formulation for considering
collision candidates is made and the existing method is reformulated. The method was examined by comparing
its results with the results of traffic simulations. This examination suggests the rationality of the model.

1 INTRODUCTION water way considering the shape of ships (Pedersen


2010). He used a closed region around a ship which
As Table 1 shows collision between ships has been is defined in the same way as Fujii defined. He
still very numerous and hazardous casualty to life and also used “collision diameter” defined by Fujii in his
environment globally. As the results of collision water formulation.
ingress and capsizing occur, persons on board were Kaneko developed a method that enables to esti-
killed and serious water pollution was caused. Most mate number of collision candidate analytically under
collision accidents are considered to be caused in such the assumption that a ship appears at the circular border
a manner that after occurrences of collision candidates and progresses in the circular area with random direc-
which are situations in which related ships collide with tion (Kaneko 2004). Kaneko also developed a method
each other without evasive action, responses of related that enables to estimate the number of collision candi-
ships to the situations are failed. The probability of dates in the area of unique ship density by defining the
failing evasive action is called “causation probability” ship’s shape as rectangle (Kaneko 2007). Kaneko esti-
(Pedersen 2010). This approach for modeling collision mated the number of collision candidates in real sea
is regarded as common way to consider collision now. area in Japan and estimated the collision frequency
Fujii proposed this approach first (Fujii 1970). The of the area using the estimated number of collision
case that while no collision candidates occur sudden candidates obtained by application of the method and
turning of a ship to the near ship occur and shortly number of collision accidents which were written in
after collision between those ships occur, rarely occurs. the casualty statistics from the Japan Coast Guard
However this case is out of consideration in this paper. (Kaneko 2007).
The important thing in judging collision candidate Montewka developed a collision model in which
is to model the collision situation between two ships, a ship is represented as a disc and MDTC (min-
that is, to model geometrical collision situation consi- imum distance to collision) is defined instead of
dering ship shape. the abovementioned “collision diameter”. To deter-
Fujii considered collision situation between ships mine MDTC value ship dynamics was considered
which sail in mutually crossed waterways. He defined (Montewka 2010, 2012). He criticized collision mod-
closed region the border of which is a trajectory of the els by Fujii, Pedersen and Kaneko as not considering
center point of an other ship obtained by sliding the ship dynamics. He also suspected that their model may
other ship keeping the progress direction of it along not fully reflect the real interaction between ships at
the edge of an own ship and keeping it contact with close quarters (Montewka 2010). However rationality
the own ship (Fujii 1970). He also defined the “colli- of their model can be supported by the abovementioned
sion diameter” which is the length of projection of the definition of collision candidates. The result of inter-
region to the direction of the relative velocity vector. action between relevant ships at close quarters may
Pedersen developed the method that enables to affects causation probability along as ship dynamics
estimate number of collision candidate in crossing affects it (Pedersen 2010). Therefore collision can-
area between two water ways which have probabilis- didate based on geometrical collisional situation is
tic distribution of ship’s position laterally along each essential for analysis of occurrence of collision.

27
Table 1. Significant casualties from 2000 to 2011 (IHS
Fairplay casualty database).

Volume of
spilled
Occurrence Number of liquids
Casualty number fatalities (Kilo Liter)

Collision 3,181 421 55,353.0


Contact 1,357 47 19,169.2
Foundered 582 6,622 11,024.6
Fire/Explosion 1,123 1,751 25,042.0
Hull/machinery 5,030 40 6,832.1
damage
War loss etc. 30 43 352.8
Missing 12 72 0.0 Figure 1. Expression of crossed waterways.
Grounding 3,246 1,127 59,040.3
Miscellaneous 60 22 12.0

Here the shape of an IA is assumed to be a rectangle


and a pair of edges which are parallel is parallel to one
However in case of small crossing angle to the of the crossing waterways. The complementary prob-
extent that two crossing waterways are almost paral- lem to “Problem 1” is the problem to count the number
lel, such as a juncture of river branches and closely inside an IA which does not include the whole crossing
located waterways near a coast, the models by Fujii and (Figure 1(b)). The problem is called “Problem 2”. A
Pedersen cannot be applied to for estimating number of case that a whole IA is included in the crossing area is
collision candidates. In such crossing area the length called Problem 2C. As the methods used to solve Prob-
of the crossing is relatively so long that collision candi- lem 2C can be applied to the other cases of Problem
dates are likely to happen even though such a crossing 2, only methods for solving Problem 2C are discussed
is narrow. Here the crossing angle is categorized into here. In Problem 1 all considered ships pass through
two types by its sharpness, and an applicable method the crossing perfectly. In Problem 2, in an IA there exist
for estimating number of collision candidates to each ships which do not pass through the crossing perfectly.
type is discussed. The method for estimating the num- If the angle between crossed waterways is large, that
ber of collision candidates in a crossing with small is, near π/2 radian, it is easy to set an IA as it includes
angle is newly developed and is an improvement on a whole crossing area. On the other hand if the angle is
method for estimating number of collision candidates. small, that is, near 0 or π radian, it becomes difficult to
set an IA in such a manner. This indicates that in case
of a crossing with a small angle Problem 2 is essential.
2 HOLISTIC FORMULATION FOR
The basic process of calculating the number of col-
ESTIMATION OF NUMBER OF COLLISION
lision candidates is to count colliding ships in the other
CANDIDATES AT CROSSING AREA
waterway while a ship in the waterway of upright posi-
tion sails in the considered time. The former ship is
In the followings the shape of a ship is approximated to
called the “other ship” and the latter ship is called
be rectangular, the length and the width of which are
the “own ship” here. After that by multiplying the esti-
the length and the breadth of a ship (Kaneko 2007).
mated number of colliding ships in the other waterway
One of crossed waterways is set upright so that the
with the number of ships in the waterway of upright
other waterway crosses it from the left to the right.
position which sail in the crossing in the considered
Because the case that the other waterway crosses the
time the number of collision candidates in the crossing
vertically set waterway from the right to the left can
can be obtained.
be considered to be a mirror image of the crossing
Prior to the formulation of the problems the follow-
situation from the left to the right, the handling does
ing definition of symbols are made. In the definitions
not lose generality.
suffix i denotes own (i = 1) or other ship (i = 2).
The crossing angle is defined as shown in Figure 1.
Li : Length of a ship
The crossing angle of opposite directional waterway
Bi : Breadth of a ship
is 0 radian and that of co-directional waterway is π
radian by this definition. At the edge of each water- Vi : Velocity vector
ways a ship appears following Poisson distribution Vi : Absolute value of velocity vector.
mean appearance rate of which is λ1 (x1 ) or λ2 (x2 ) V2 − V1 : Relative velocity vector of an other ship
respectively. to an own ship.
Here an inspection area (IA) is set to count the |V2 − V1 |: Absolute value of the relative velocity
number of collision candidates in the crossing of two θ: Crossing angle of waterway 2 to waterway 1
mutually crossed waterways. If an IA includes the defined in Figure 1
whole crossing, the problem to count a number of θ−V1 ·V2 −V1 : Angle between V2 − V1 and reverse
collision candidates is called Problem 1 (Figure 1(a)). direction to the progress direction of an own ship.

28
Figure 2. Collision area (CA) and collision polygon (CP).

Figure 3. Relation between CA and crossed water way and


collision segment (CS) at problems. Figure 4. Case division of Problem 1.

2.1 Problem 1
For simplifying explanatory figures below a filled This problem includes four cases considering an angle
rectangle () is used instead of θ−V1 ·V2 −V1 . between relative the velocity vector V2 − V1 and the
θV2 ·V2 −V1 : Angle between V2 − V1 and the progress velocity vector of an own ship V1 . The case division is
direction of an other ship. Similarl to the treatment illustrated in Figure 4.
of θ−V1 ·V2 −V1 , a filled triangle () is used instead of The positions of other ships on a CS which corre-
θV2 ·V2 −V1 . spond to the points on the border of a CP are illustrated
An other ship collides with an own ship when she in figures from Figure 5(a) to (d) at every cases. Ships
comes into the polygonal area, the boundary of which of both ends of a CS is only tangent to a CP. Although
is the trajectory of a center point of an other ship the positions of the other ships at the border of a CP
obtained by sliding the other ship while keeping the differ at every cases, the width of both lines which con-
progress direction of an other ship along the edge of nect both ends of a CS and corresponding points on
an own ship and keeping it contact with the own ship. the border of CP is expressed by the same equation,
The polygonal area is called the “Collision polygon that is, Equation (1). The width has been called the
(CP)” (Figure 2). Other ships which collide with an “collision diameter” until now (ex. (Pedersen 2010)).
own ship during duration T locates in the area shown
in Figure 3. The both edges of the area are parts of a
CP and both sides are line segments which is tangent
to the CP of both sides of the area. The length of the
line segments of both sides of the area is |V2 − V1 | · T
and the slant is the slope of a vector of V2 − V1 . From law of sines Equation (2) holds.
The area is called “Collision area (CA)”. Ships on
the line segment which is obtained after cutting a sail-
ing line of other ships by CA collide with the own ship.
The line segment is called “Collision Segment (CS)”
(Figure 3). Problem 1 is the case that both boundaries
of waterway 2 intersect the side lines of CA. Prob- Then Equation (1) is transformed into Equation (3)
lem 2C is the case that whole CA is included inside
of waterway 2. There exists the case that only a part
of CA is included in waterway 2. As stated above the
case can be dealt with using the methods used to the
above case. Therefore the case is not dealt with here.

29
Equation (3) is the same form as the length of colli-
sion diameter in (Pedersen 2010). The length of CS
(LCOL ) is related to collision diameter (DCOL ). The
relation is expressed as equation (4).

The numbers of ships which collide with an own


ship during duration T while the own ship passes
through the crossing area between waterways can be
estimated by integrating numbers of ships on a CS
throughout a CA. As a ship is assumed to appear at
the edge of waterway following Poisson distribution,
the number is obtained if the sailing time in CS is esti-
mated. The time is obtained to divide the length of CS
by the velocity of an other ship (V2 ).
Then the mean value of number of ships in CS(NS )
is expressed by Equation (5).

Next the parameter of Poisson distribution (λi (xi )),


that is, the mean value of ship appearance rate, can be
expressed as λi (xi ) = i · ki (xi ) using the following
values. In the definitions suffix i denotes own (i = 1)
or other ship (i = 2).
i : Number of ships which appear at the end of a
waterway during unit time
ki (xi ): Probabilistic distribution function (PDF) of
position where ship appear
Then in the case that all ships pass through the cross-
ing between waterway 1 and 2 number of collision
candidate (NCOL ) is obtained by Equation (6) during
duration T that is much longer than the time for passing
through the crossing.

Equation (6) is the same form with that of (Pedersen


2010). As Equation (6) does not include any function
of x1 and x2 , it is simplified as Equation (7).

Figure 5. Collision segment (CS) and collision diameter in


cases of Problem 1.

2.2 Problem 2C be S/(2*DCOL ). S means surface of a CP. S is expressed


by Equation (8).
CA defined above has edge parts with a complicated
shape. Therefore in order to simplify the case division
needed to estimate number of collision candidates at
this problem, the shape of edges of the CA is trans-
formed into a rectangle the width of which is DCOL
and the length(d) defined in Figure 6 is calculated to

30
Figure 6. Simplified collision area (CA) used in
Problem 2C.

This transformation of edges of a CA has a merit


that transformed edges becomes the right shape at the Figure 7. Relation between CS and CA in Case 1.
crossing angle is 0 and π.
This problem is divided into several cases by a cross- E5 is a point of intersection of E1 E2 and a line which
ing angle and the difference of edges of a CA which a passes through E3 and is parallel to V2 . E6 is a point
CS crosses. In the following an IA is defined in front of intersection of line XW2 and E3 E5 . E7 is a point
of an own ship inside the crossing and it is assumed of intersection of line XW2 and a line which passes
that an own ship appears at the lower edge of the IA. through E2 and is parallel to V2 . E8 is a point of inter-
DW1 is defined as the length of IA. Then the time that section of E4 E3 and a line which passes through E1 and
an own ship passes through the IA is DW1 /V1 . is parallel to V2 . E9 is a point of intersection of line
In the following sections several auxiliary lines and XW2 and a line which passes through E4 and is paral-
points are defined in explanatory figures to obtain lel to V2 . E10 is a point of intersection of line XW2 and
an equation which express the number of collision E1 E8 . E11 is a point of intersection of line XW1 and
candidates. As explanation of those lines and points a line which passes through E1 and is perpendicular
consumes so much space that the explanation is made to XW1 . E12 is a point of intersection of line XW1 and
only in Figure 7. Auxiliary lines and points in other E1 E4 . E13 is a foot of the perpendicular to E1 E4 from
figures can be understood easily by analogy with the PS1 which is an appearance position of an own ship.
explanation of Figure 7. E14 is also a foot of the perpendicular to line XW2 from
E1 . H is the distance between line XW1 and XW2 . x1
2.2.1 Case 1: θ ≈ 0 and a CS crosses both a short is distance between E6 and a point of intersection of
edge and a long edge or a CS crosses short the line XW2 and a line which is parallel to E3 E5 and
edges of both sides of a CA intersects E1 E2 and E2 E3 . y1 is length of a line seg-
This case corresponds to the following inequality. ment obtained by cutting the line which is parallel to
E3 E5 and intersects E1 E2 and E2 E3 by E1 E2 and E2 E3 .
x2 is distance between E9 and a point of intersection of
the line XW2 and a line which is parallel to E1 E6 and
intersects E1 E4 and E4 E3 . y2 is length of a line seg-
ment obtained by cutting the line which is parallel to
In Figure 7 the rectangle E1 E2 E3 E4 is a CA. PS1 E3 E5 and intersects E1 E2 and E2 E3 by E1 E2 and E2 E3 .
is appearance position of an own ship. The own ship Then length of E1 E8 , y1 and y2 are obtained below.
appears at an optional position on the line XW1 , coor-
dinates of which is denotes as (xW1 , 0) following the
Poisson distribution parameter of which is 1 k1 (xW1 ),
and she sails upward. The other ship appears at an
optional position on the line XW2 , coordinates of which NCOL (T) of this case is estimated by Equation (9)
is denoted as (xW2 , H) following the Poisson distri- using Z defined below.
bution parameter of which is 2 k2 (xW2 ) and sails
diagonally downward with crossing angle θ. In case
that the own ship sails from its original position to the
other edge of the IA, the length of a CS cut by the CA,
which is included in the integrand for obtaining NCOL
can be derived from Figure 7. Upper and lower limits
of the domains of integration also can be derived from
Figure 7.

31
The upper and lower limits of the domains of inte-
gration, that is E6 (xW2 ), E7 (xW2 ), E9 (xW2 ), E10 (xW2 ),
are obtained below.

Figure 8. Relation between CS and CA in Case 2.

2.2.3 Case 3: θ ≈ π and 4 cases defined by


overtaking relation between an own and an
As integral forms of NCOL in the cases after this case other ship and crossing relations between a
are all the same except integrand “Z” for saving space CS and edges of a CA
the integral forms are not described in the following This case is rather complicated. In this case relation
cases. Only integrand “Z” is shown. between velocity of an own ship and an other ship the
θ = 0, that is the opposite directional is a special situation that an other ship overtakes an own ship and
one of this case. In this case as E9 and E10 are same the opposite situation exists if there exists probabilistic
point, E6 and E7 are also same point, y1 and y2 cannot distribution of velocity of ships. In the case that veloc-
be defined. Moreover the following relations hold. ity’s mean value of an own ship and that of an other
ship are nearly equal, those opposite situations occur
frequently. In the case that an other ship overtakes an
own ship CA is made at left lower of an own ship,
Therefore at θ = 0, integrand “Z” is expressed by in the opposite case CA is made at left upper of an
Equation (10). own ship. Moreover in every cases there are two cases
which are case 1 and case 2 of Problem 2C. Therefore
the four cases explained in the followings should be
considered.

2.2.2 Case 2: θ ≈ 0 and a CS crosses both a short 2.2.3.1 Case 3.1: θ ≈ π and an own ship overtakes
edge and a long edge or a CS crosses long an other ship and a CS crosses both a short
edges of both sides of a CA edge and a long edge or a CS crosses short
This case corresponds to the following inequality. edges of both sides of a CA
In this case velocity of an own ship and that of an other
ship have the relation below.

Similar to Case 1 length of CS cut by CA, which


is included in integrand for obtaining NCOL can be In addition following inequality holds.
derived from Figure 8. Upper and lower limits of the
domains of integration can be derived from Figure 8.
Then length of E1 E15 , y1 and y2 are obtained below.

In Figure 9 E1 E2 E3 E4 is a CA. PS1 is appearance


In this case, integrand “Z” is expressed by position of an own ship. QS1 is the relative position of
Equation (11) the center point of other ships which collide with the
own ship at the final position of the own ship. An own
ship appears at an optional position of line XW1 being
followed to PDF of an own ship’s appearance and sails
upward. An other ship appears at an optional position
of line XW2 being followed to PDF of an other ship’s

32
Figure 9. Relation between CS and CA in Case 3.1.
Figure 10. Relation between CS and CA in Case 3.2.

appearance, and sails diagonally upward at crossing 2.2.3.2 Case 3.2: θ ≈ π and an own ship overtakes
angle θ. an other ship and a CS crosses both a short
In case that an own ship sails from appearance posi- edge and a long edge or a CS crosses long
tion to the opposite edge of the IA, length of a CS cut edges of both sides of a CA
by the CA, which is included in integrand for obtaining Similar to Case 3.1 of Problem 2C, velocity of an own
NCOL can be derived from Figure 9. Upper and lower ship and that of an other ship have the relation below.
limits of the domains of integration can be derived
from Figure 9. Length of E1 E8 , y1 and y2 are obtained
below.
In addition following inequality holds.

Then in this case, integrand “Z” is expressed by The length of a CS cut by a CA, which is included
Equation (12). in integrand for obtaining NCOL can be derived from
Figure 10. Length of E2 E5 , y1 and y2 are obtained
below.

Then in this case, integrand “Z” is expressed by


θ = π, that is the co-directional is a special one of Equation (14).
this case. In this case as E9 and E10 are same point,
E6 and E7 are also same point, y1 and y2 cannot be
defined. Moreover the following relations hold.

2.2.3.3 Case 3.3: θ ≈ π and an other ship overtakes


an own ship and a CS crosses both a short
Then in this case, integrand “Z” is expressed by edge and a long edge or a CS crosses short
Equation (13). edges of both sides of a CA
In this case velocity of an own ship and that of an other
ship have the relation below.

33
In addition following inequality holds.

In this case a CA locates lower left of an own ship


and there would be the case that the line where other
ships appear crosses the CA. To deal with this case a
treatment explained in the followings to estimate coor-
dinate value of the appearance position of an other ship
applied to. The line where other ships appear is named
as XW2 and a position where other ships which collide
with the own ship appear is named using a symbol with
a dash, that is, E’6 , E’7 , E’9 , E’10 .
The length of a CS cut by the CA, which is included
in integrand for obtaining NCOL can be derived from
Figure 11. E’6 (xW2 ), E’7 (xW2 ), E’9 (xW2 ), E’10 (xW2 ),
length of E4 E5 , y1 and y2 are obtained below.
Figure 11. Relation between CS and CA in Case 3.3.

Then in this case, integrand “Z” is expressed by


Equation (15).

θ = π, that is the co-directional is a special one of


this case.
Figure 12. Relation between CS and CA in Case 3.4.
In this case as E9 and E10 are same point, E6 and
E7 are also same point, y1 and y2 cannot be defined.
Moreover the following relations hold. In addition following inequality holds.

Then in this case, integrand “Z” is expressed by


Equation (16). A CA locates also lower left of an own ship and there
would be the case that the line where other ships appear
crosses the CA. To deal with this case a treatment
explained in the followings to estimate the coordi-
nate value of the appearance position of an other ship
2.2.3.4 Case 3.4: θ ≈ π and an other ship overtakes applied to. The line where other ships appear is named
an own ship and a CS crosses both a short as XW2 and a position where other ships which collide
edge and a long edge or a CS crosses long with the own ship appear is named using a symbol with
edges of both sides of a CA a dash. E’6 (xW2 ), E’7 (xW2 ), E’9 (xW2 ) and E’10 (xW2 ) are
Similar to case 3.3 velocity of an own ship and that of obtained by same equations of Case 3.3.
an other ship have the relation below. The length of a CS cut by the CA, which is included
in integrand for obtaining NCOL can be derived from
Figure 12.

34
E4 E19 , y1 and y2 are obtained below.

Then in this case, integrand “Z” is expressed by


Equation (17).

3 TRAFFIC SIMULATION

Assuming that own ships and other ships appear at


edges of Waterway 1 and 2 being followed to Poisson
distributions which are defined at each waterway
respectively, and that the appearance positions of own
ships and other ships are followed to normal dis-
tributions which are also defined at each waterway Figure 13. Target area of simulation at Problem 1.
respectively, ship traffic simulations were carried out
in both Problem 1 and Problem 2C. The formula-
tions made in chapter 2 were examined comparing the
results with corresponding results by simulations.
Since probabilistic distribution of ship appear-
ance assumed to be normal distribution, ends of ship
appearance line segment are meaningless. Important
values are the center of the line segment and standard
deviation.
Collision between ships is judged by finding the
intersection of ships which sail in mutually different
waterways.

3.1 Problem 1 Figure 14. A snapshot of a traffic simulation of Problem 1


(crossing angle: 1/3 π).
Conditions of simulation are as follows.
Duration: One hour
This means that counting number of collision was
continued until all own ships which appear during MATHEMATICA. In the column “Simulation” the
continuous one hour had passed through the crossing number obtained by simulation was shown.
area. It can be said that values by the model is near to cor-
Crossing angle: π/4, π/3, π/2, 2π/3, 3π/4 responding value by simulation. However number of
Number of times of simulation: 10 (at each crossing times of simulations is too few to judge whether there
angle) is significant difference between the theoretical model
Additional conditions were set in Table 2 to every and the corresponding simulation or not statistically.
crossing angle.
Figure 13 shows the target area used in the sim-
ulation for Problem 1. IA was set as it included the
3.2 Problem 2C
crossing area between Waterway 1 and 2. Own ships
and other ships sail until they reach their terminal lines. Simulation condition is as follows.
Figure 14 is a snapshot of a traffic simulation Duration: One hour
of Problem 1, where crossing angle is 1/3π. Filled This means that counting number of collision
rectangles are ships in Waterway 2 which collide with continued until all own ships which appear within
ships in Waterway 1. continuous one hour had passed through IA.
Table 3 shows the results obtained by the theoretical Crossing angle: 0, π/50, π/20, (45/50)π, (49/50)π, π
model and by the simulation. Number of times of simulation: 10 (at each crossing
In the column “Theory” the values which were angle)
obtained by models was shown. The integra- Additional conditions were set in Table 4 to every
tion of Equation (7) was carried out by using crossing angle.

35
Table 2. Traffic simulation condition (Problem 1). Table 4. Traffic simulation condition (Problem 2C).

Items Waterway 1 Waterway 2 Waterway 2

Ship appearance rate Items Waterway 1 θ ≈ 0 θ≈π


Type of distribution Poisson Poisson
Mean (1/sec) 1/60 2/60 Ship appearance rate
Type of distribution Poisson Poisson Poisson
Ship appearance position Mean (1/sec) 1/60 2/60 2/60
Type of probabilistic Normal Normal
distribution Ship appearance
position
Coordinates of center (1450, 0) (0, XW2min + 250)
Type of distribution Normal Normal Normal
point
Coordinates of (1450, 0) (1400− (1400−
standard deviation (m) 87.5 67.5 center point 2000 ∗ tan(θ), 500 ∗ tan
Velocity 4300) (π − θ), 0)
Type of distribution Normal Normal standard 87.5 200 + 250 ∗ 200 + 250∗
mean (kt) 10 20 deviation (m) tan(θ) tan(π − θ)
standard deviation (kt) 1 2 Inspection Area (IA)
Ship length Lower left (1100, 300) (1100, 500)
Type of distribution Normal Normal Upper right (1800, 1300) (1800, 500)
mean (m) 250 150 Velocity
standard deviation (m) 25 15 Type of distribution Normal Normal Normal
mean (kt) 10 20 20
Ship Breadth
standard 1 2 2
Type of distribution Normal Normal
deviation (kt)
mean (m) 40 30
standard deviation (m) 4 3 Length of ship (m) 250 150 150
(Constant)
π 
*XW 2MIN = 1000 + 1700 ∗ tan 2
−θ Breadth of ship (m) 40 30 30
(Constant)

Table 3. Number of collision candidates by theoretical


model and by simulation (Problem 1).

Simulation
Crossing
angle Theory Average Maximum Minimum

π/4 170.7 168.0 210 145


π/3 162.8 162.7 220 118
π/2 147.3 153.6 189 120
2π/3 138.9 145.6 184 124
3π/4 143.0 139.9 210 85

In this problem integration was carried out in the


manner that every used random valuable is digi-
tized by suitable interval and summarized all number
of collision candidate obtained at the combination
of valuables. For saving time of integration, length
and breadth of a ship are assumed to be constant
respectively and MATHEMATICA was not used.
Figure 15 shows the target area used in the simula-
tion for Problem 2C where crossing angle (θ) is close Figure 15. Target area of simulation at Problem 2C where
crossing angle (θ) is close to zero.
to zero. Figure 16 also shows the target area used in the
simulation for Problem 2C where crossing angle (θ) is
close to π. IAs are hatched rectangles in these figures.
Own ships and other ships sail until they reach their Results are summarized in Table 5.
terminal lines. Similar to Problem 1 it is said that values from
Figure 17 is a snapshot of a traffic simulation of theory is near to the corresponding value from sim-
Problem 1, where crossing angle is 45/50π. Filled rect- ulation. However simulation number is too few to
angles are ships in Waterway 2 which collide with ships judge whether there is a significant difference or not
in Waterway 1. statistically.

36
Table 5. Number of collision candidates by developed
method and by simulation (Problem 2).

Simulation
Crossing
angle Theory Average Maximum Minimum

0 90.7 84.3 106 55


1/50π 82.9 74.6 111 50
1/20π 66.7 63.3 96 44
45/50π 80.3 71.7 103 56
49/50π 54.8 46.0 72 20
π 69.2 56.4 76 46

existing method already. Problem 2 has not been dealt


with by existing method. Therefore it can be said that
the developed method for Problem 2 can be said to
be an improvement added to the existing method for
Problem 1.
Moreover at every sub-problem those methods were
verified by traffic simulation. For establishment of
those methods for estimation of number of the col-
lision candidates statistical verification is necessary.
Figure 16. Target area of simulation at Problem 2C where For this purpose a lot of traffic simulations and field
crossing angle (θ) is close to π radian. measurement are necessary.

REFERENCES
Fujii, Y.; Yamanouch, H. & Mizuki, N. 1970. On the fun-
damentals of marine traffic control. Part I: Probabilities
of collision and evasive actions. Electronic Navigation
Research Institute Papers 2: 1–16.
Kaneko, F. 2004. Effectiveness of separation scheme for
prevention of collision by diminishing ships’ encounter
probability, Proceedings of the 3rd international con-
ference on collision and grounding of ships (ICCGS):
211–220. Izu, Japan.
Kaneko, F. & Hara, D. 2007. Estimation of dangerous
encounters’ number from observed ship trajectories.
4th international conference on collision and grounding
of ships. Hamburg University of Technology (ICCGS).
Schiffbautechnische Gesellschaft. 187–19. Hamburg,
Figure 17. A snapshot of a traffic simulation of Problem 2C Germany.
(crossing angle: 45/50π). Pedersen, P.T. 2010. Review and application of ship collision
and grounding analysis procedures. Marine Structures 23:
241–262.
Montewka, J.; Hinz, T.; Kujala, P. & Matusiak, J. 2010.
4 CONCLUSION Probability modeling of vessel collisions. Reliability
Engineering and System Safety 95: 573–589.
The existing method for estimation of the number Montewka, J.; Goerlandt, F. & Kujala, P. 2011. Determina-
of collision candidates at the crossing of two water tion of collision criteria and causation factors appropriate
ways was reexamined and a holistic formulation for to a model for estimating the probability of maritime
the estimation of the collision candidates was made. accidents. Ocean Engineering 40: 50–61.
In the formulation the problem of the estimation of
the number of collision candidates was divided into
two sub-problems, that is Problem 1 and Problem 2
by the relation between an inspection area (IA) and
the crossing. Problem 1 has been dealt with by the

37

You might also like