International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 4
Number 3
September 2010
261
1 INTRODUCTION
Although modern watercrafts have been equipped
with advanced facilities, such as satellite navigators,
anti-collision radars, electronic charts, and automatic
cabins, etc., the occurrence of shipping accidents is
still frequent. Therefore, the IMO organization and
many other organizations pay more attention to the
shipping safety. According to IMO organization’s
analysis results, there are three main reasons result-
ing in shipping accidents in recent years:
1 the failure of equipments on watercrafts;
2 the influence of navigation environments and
conditions;
3 and, human factors.
The proportion of these three factors to the whole is
shown in Fig.1, from which it can be seen that up to
eighty percent of shipping accidents are directly or
indirectly caused by human error. Under the compli-
cated circumstance, navigators lacking of watch
have poor capability of dealing with emergency,
poor qualities and little sense of duty, as well as
cannot master traffic rules on the sea well. Hence,
misoperation will occur. In order to guarantee the
safety of shipping, we should pay more attention to
human factors where not only the quality of naviga-
tors and their ability of dealing with emergency
would be developed, but also the latest modern tech-
nology for prediction of shipping accidents and
blocking the wrong order or misoperation would be
adopted. Consequently, a shipping accident forecast
model is developed in this paper. Furthermore, a
method of the shipping accident prediction and con-
trol is studied based on the proposed model.
2 ANALYSIS OF THE REASONS IN SHIPPING
ACCIDENTS
The shipping system is comprised of watercrafts,
human and the navigation circumstance. Due to the
particularity of carriers and conditions, the charac-
teristic of shipping systems which are complicated
systems is multi-category, multi-layer, multi-
attribute and multi-rule. Shipping accidents can be
classified into the following types, collision, ground-
ing, striking, heavy weather, fire & explosion,
foundering, missing, and engine failure, etc. Risk
during navigation mainly results from three factors:
a) collisions with other objects (static or moving), b)
the change of external environments (such as ty-
phoons, tidal waves, fogs, and other disasters), and
c) wrong orders sent by navigators (including
misoperations). These three factors of the system al-
so have many components which are shown in Fig.1.
In recent years, new challenges encountered in ship-
ping systems are mainly as follows:
Study on Shipping Safety Strategy Based on
Accident Forecast Model
X.Y. Meng, Y.M. Bai & X.J. Han
Dalian Maritime University, Dalian, China
ABSTRACT: The factors which would cause shipping accidents are analyzed in detail and a model which can
forecast shipping accidents is studied in this paper. During navigation, all the factors are integrated and calcu-
lated in this model which then estimate and speculate on the risk degree of collision for the own ship. Finally,
the risk level and the possibility of shipping accidents can be forecasted in real-time. The proposed accident
forecast model can estimate the possibility of collision with other ships or objects in a specific domain.
Meanwhile, the external environment such as weather, stream, etc. is taken into account in the model. Be-
sides, the validity of navigators’ orders can also be evaluated in the model which consequently can forecast
different kinds of shipping accidents effectively in that most of the factors which cause shipping accidents
have been involved in the proposed model. With the accident forecast model, the shipping safety would be
improved greatly. A practical example demonstrates the effectiveness and superiority of the proposed strate-
gy.
262
Figure.1 The influence factors of shipping accident
1) With the trend that ships become larger and larg-
er, as well as more and more rapid, the inertia of a
ship becomes larger, and therefore it is more dif-
ficult to manipulate the ship.
2) With the world economy incorporating and in-
creasing, the quantity of ships proliferates. To-
gether with the progress of ocean oil field explor-
ing and sea culture, the navigation density inshore
becomes higher. Hence, high risk would be taken
when a ship navigates in the inshore area or on
dense lane.
3) The abnormal change of the weather such as fogs
and typhoons, etc., critically influences the navi-
gation condition.
In order to cope with these problems, the technology
and techniques for shipping safety are required to be
studied and developed urgently. Especially, the hu-
man factor should be highlighted. In this paper, a
shipping accident forecast model is proposed to
avoid and control various accidents.
3 THE SHIPPING ACCIDENT FORECAST
MODEL
3.1 Structure of the model
Nowadays, there are many kinds of models describ-
ing the ship motion. Among them, models correlated
to the shipping safety are mainly: 1) OD traffic flow
model (Fuji J. et al. 1971), 2) ship field model (Da-
vis P V. et al. 1982, Goodwin E.M. 1975), and 3)
DCPA (distance of close point of approaching) and
TCPA (time of close point of approaching) model
(Zheng Zhongyi et al. 2000). These models describe
various characteristics of the shipping safety from
different points of view. They have some advantages
and are applied to large scales of areas. However,
they cannot essentially solve the problems of the
shipping safety. In this paper, a shipping accident
forecast model which integrates each kind of factors
influencing the shipping safety is proposed. As the
output of the model, the risk degree is the concept of
the possibility that accident will occur.
The shipping accident forecast model can be ex-
pressed as follows:
njitTdTT
HMUUUnW
,
),()( =
(1)
where,
)(nW
——the output of the forecast model which is
defined as risk degree;
n
H
——the evaluated influence degree of naviga-
tion environments;
T
U
——the risk degree of collision with encounter-
ing targets;
ji
M
,
——evaluation results of the operation instruc-
tion (telegraph orders and rudder orders).
The output of the accident forecast model is de-
termined by three items which correlates to three
main aspects causing shipping accidents respective-
ly. The first item in the model is
),(
tTdTT
UUU
,
which interprets the encounter probability and the
risk degree of collision. The second item is the eval-
uation to the validity of orders sent by navigators.
The third item is the degree of influence on the ship-
ping safety while external conditions vary. The out-
put
)(nW
∈ [0, 1] suggests that the ship has no dan-
ger while
)(nW
=0 and the ship is in danger while
)(nW
≠0. The larger the output
)(nW
is, the higher
the risk degree is.
The inputs of the model mainly come from the
scanning information of the ARPA, other navigation
operation instructions, parameters of velocity and
course, and other related information from sensors
(wind velocity and ship draft, etc.). Integrating this
information, the forecast model evaluates the risk
degree in real time.
3.2 Risk degree of collision
The risk degree of collision with encounter targets
T
U
involves the space collision risk
T
d
U
and the
time collision risk
T
t
U
. The space collision risk
mainly includes the
DCPA
, ship fields, the fuzziness
of domain boundary, the orientation of encounter
targets, observation errors in the
DCPA
, etc. Ac-
cording to the velocity and course of the own ship
263
and encounter objects, the shortest encounter dis-
tance between them is
)sin(
παϕ
−−•=
TRT
RDCPA
. After the safety en-
counter domain
1
d
and the safety passing distance
2
d
are determined, the fuzzy set
T
d
U
of the space colli-
sion risk can be obtained. The membership function
T
d
u
of
T
d
U
is defined as follows:
>
≤≤
−
−
<
=
2
21
2
12
2
1
|| if ,0
|| if ,
||
|| if ,1
)(
dDCPA
dDCPAd
dd
DCPAd
dDCPA
DCPAu
dT
The time collision risk mainly expresses the rela-
tive velocity, distance, velocity ratio between two
ships, the length of own ship, and maneuvering per-
formance, etc. According to the relationship between
encounter and movement of ships, the encounter
time between the own ship and targets is given by
RTRT
vRTCPA /)cos(
παϕ
−−•=
(2)
After the extreme time
1
t
for sending a rudder order
and the time
2
t
for ensuring the relative safety dis-
tance between two ships are determined，the verify-
ing domain of
TCPA
is
t
U
and the fuzzy set of the
time collision risk is
tT
U
, of which the correspond-
ing membership function
tT
u
is defined as follows:
>
≤<
−
−
≤
=
2
21
2
12
2
1
if ,0
if ,
if ,1
)(
tTCPA
tTCPAt
tt
TCPAt
tTCPA
DCPAu
tT
.
The collision risk between ships is the synthesis of
the space collision risk and the time collision risk. In
the domain
U
, the ship collision risk is a set of
T
U
,
and we have
iTdTT
uuu ⊕=
(3)
The above-mentioned synthesis operator “
⊕
” means
that,
So long as
dT
u
=0 or
iT
u
=0, we have
0=
T
u
. Other-
wise, if
0≠
dT
u
and
0≠
iT
u
,
)max(
, iTdTT
uuu =
.
3.3 Determination of the impact item of
environment conditions
The navigation environment factors involve the
weather, hydrology and lane conditions. For the sake
of simple computations, the environment influence
parameter
n
H
(influence degree of navigation envi-
ronment) is mainly determined by the wind velocity
and the wind direction measured by wind gauges,
and determined by rocking parameters which are de-
rived from the ship’s draft of each shipboard. The
function is given by
)(
maxmax
l
l
f
f
H
nn
n
⊗=
Where
n
f
is the wind velocity,
max
f
is the maximal
wind power,
n
l
is the rocking height,
max
l
is the max-
imal rocking height.
max
f
and
max
l
can be approxi-
mately set according to actual navigation conditions.
The value of
n
H
∈[0, 1] is the maximum of
n
n
f
f
max
and
n
n
l
l
max
, which means that the worse the environ-
ment is, the more the value of
n
H
is close to 1. It is
obvious that the environment influence parameter
n
H
makes a great impact on the risk degree
)(nW
.
3.4 Assessment of navigators’ orders
Most shipping accidents are caused by improper or-
ders sent by navigators. Hence, as the term for as-
sessing the validity of orders,
ji
M
,
is included in the
forecast model. The principle of how to get the value
of
ji
M
,
is shown in Fig. 2.
Figure.2 The choice of value M
i,j
① Expert system of the order assessment is stored in
the system. Each order sent by navigators will be
analyzed whether it is reasonable according to basic
operation rules and emergency operation rules. If the
order violates the rules, then
ji
M
,
=1. Fig.2 The
choice of the value
ji
M
,
output
)(nW
equals to ze-
ro.
264
② If the order is evaluated to be valid by the expert
system, the output
)1( +nW
, which is the output on
the time of n+1 after the order is sent, should be cal-
culated and compared with the output
)(nW
on the
time of n. The variety of
)(nW
will be obtained
through the function
)()1( nWnWW −+=∆
. If
W∆
is positive, it implies that the risk degree is increas-
ing, then
ji
M
,
=
W∆
. Otherwise,
ji
M
,
=0.
4 PRACTICAL EXAMPLE
A practical example of the shipping accident fore-
cast model will be illustrated as follows. A ship is
navigating with the course of 60° and the velocity of
14Kn, as shown in Fig. 3.
Figure.3. The navigating condition of own ship
In the definite domain, the own ship may encounter
with other five ships which are marked as G
1
, G
2
,
G
3
, G
4
and G
5
. The moving condition of ships is
listed in Table 1.
Table 1. The moving condition of other ships in the domain of
the own ship
object
G
1
G
2
G
3
G
4
G
5
course（C
0
）
velocity（V）
distance（D）
angle of bow
（
Q
）
120
16
3.85
335
136
11
2.75
010
36
10
5.60
035
24
14
3.20
065
30
12
4.15
121
Here, DCPA, TCPA and the collision membership
function between the own ship and other ships at the
moment t0 can be deduced, it is shown in Table 2,
from which it can be seen that the collision risk de-
gree between the own ship and G
4
is the highest at
the moment t
0
,
T
U
= 0. 437. Provided that
n
H
=0
and
0=
ij
M
, we have
)(nW
=0.437. Similarly, the
output
)(nW
of the forecasting model at the time of
t
1
, t
2
, t
3
, t
4
can be calculated respectively. According
to the output of the model, navigators can take prop-
er actions to avoid shipping accidents.
Table 2. Collision membership function between the own ship
and other ships at the moment t
0
.
object
G
1
G
2
G
3
G
4
G
5
DCPA
TCPA
u
dT
u
tT
U
T
1.188
0.220
0.018
0.015
0.018
1.732
0.184
0.078
0.071
0.078
0.947
0.660
0
0.057
0
1.339
0.286
0.437
0.375
0.437
1.676
0.359
0.206
0.146
0.206
5 CONCLUSIONS
A shipping accident forecast model is proposed in
this paper, and the method of the predicting shipping
accident is developed based on the collision risk de-
gree, environmental influence coefficient and in-
struction assessment calculated in real-time. The key
technology is the integration of relevant information
of the shipping safety and the composition of the
model. Because many complicated factors are in-
volved in the shipping system, there are still some
factors which are not considered in this paper, to
name a few, evaluation rules of instructions, the reli-
ability of the sea scanning, and the influence of visi-
bility on sea. It is believed that with our tirelessly
work and continuous development of information
technology the complex shipping system can be
modeled accurately to predict and avoid shipping
accidents.
6 ACKNOWLEDGMENT
This paper is supported by “National Basic Research
Program of China” (No.2008CB417215). The Pro-
ject name is “Research on mechanism of shipping
safety and accident-forecasting theory”.
REFERENCES
Colnill, R. D. & Wignall, D. & Dand I. 2004. The application
of marine risk simulation to the “Nearcasting”and prention
of collision incidents[C],vts 2004 The 10th International
symposium on Vessel Traffic Services.
Davis, P V. & Dove, M.J.& Stockel C.T. 1982. A computer
simulation of multi-Ship Enocounters. The Journal of navi-
gation 35:347 -352.
Fuji, J & Tanaka K 1971. Traffic capacity. The Journal of nav-
igation 24:543-552.
Goodwin, E.M. 1975. A Statistical Study of Ship Domains. The
Journal of navigation 28: 329-341.
Lewison, G.R.G. 1997. The Modelling of Marine Traffic Flow
and Potential Encounters. Proceedings of International con-
fereuce on Mathematical Aspects of Marine Traffic pp.139
London
Rafal Szlapczynski 2006. A Unified Measure of Collision Risk
Derived From The Concept of A Ship Domain. The Journal
of navigation 59:477-490
Zheng Zhongyi & Wu Zhaolin 2000. ship collision avoidance
decision-making . Dalian: Dalian Maritime University.
