93
1 INTRODUCTION
A marine accident can have dire consequences such
as the loss of ships, and damage to marine
infrastructure and the environment. These accidents
haveledtoeconomiclossesandlongrecoverytimes.
In order to prevent marine accidents, safety
management is considered a significant issue in
industrialshipping
andshipnavigation.Ithas come
to light that there is a need to develop safety
evaluation methods to enhance safety during ship
navigation. Various navigational safety evaluation
methodshave,therefore,beenproposed.
Fujii (1971) and Macduff (1974) proposed a ship
collision probability model to determine a safe
navigational zone.
In these studies, the level of risk
was calculated using statistical analysis of a marine
accident and the traffic data. The obtained results
indicated the potential risk of the collision or
groundingof aship.It canbe helpful to distinguish
the hazard zones in port area. However, the
calculationresult
reliesonhistoricaldata.Whenusing
thistypeofmethodology,itisnotpossibletoreflect
real‐timenavigationalsituations.
Ship domain models have been proposed in
research regarding safety evaluation models. A
certain area around a ship, such as a circular,
rectangular, elliptical, or polygonal shape, has been
proposed.
It is to remain clear of other ships. The
shape andsize of a ship’s domain is determined by
thecalculatedsafedistanceonthebasisofstatistical
analysisofmarinetrafficdata(Goodwin,1975;Fujii,
1971),fuzzylogic(Pietrzykowski, 2008,Wang,2010),
or questionnaire results and fuzzy logic
A New Risk Evaluation Model for Safety Management
on an Entire Ship Route
S.Hwang,E.Kobayashi&N.Wakabayashi
GraduateSchoolofMaritimeSciences,KobeUniversity,Japan
N.Im
DivisionofMarineTransportationSystem,MokpoMaritimeUniversity,Korea
ABSTRACT:Inthispaper,weintroduceanewriskevaluationmodelforevaluatingthenavigationsafetyzone
foranentireshiproute.Thismodelconsiders anewalgorithmtodeterminethenavigationalsafety zonein
real‐time,andalsotakesthenavigation
officers’perceptionwhilenavigatingashipintoconsideration.Therisk
quantificationhasbeendevelopedusingaquestionnaireandincorporatedintothenewmodel.Asimulation
wascarriedoutfortheOsakabayareainordertoverifytheusefulnessoftheproposedmodel.Anewapproach
wasemployedtomonitor
thelevelofnavigationsafetyalongashiproute.Theentireshiprouteisdividedinto
smallsectionsasagriddedmatrix.Thelevelofnavigationsafetycanbequantifiedbymeansofasafetyindex
on the basis of the ship’s navigation data within a specified distance range. The results
show that the
comparison between risks identified for different sections across the entire ship route is easy, which helps
determinethenavigationalsafetyzonequickly.Thismodelisexpectedtobeabletoserveasanewtoolfor
managingsafetythroughoutanentireshiprouteareainreal‐time
inordertosupporttheportsafetyauthority
orvesseltrafficservicecenter.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 10
Number 1
March 2016
DOI:10.12716/1001.10.01.10
94
(Pietrzykowski,2009;Wang,2010).Theseresultshave
been helpful in supporting navigation officers’
decisionsofkeepingasafedistancearoundashipto
avoid collisions. However, this approach has
limitations in terms of its ability to consider
navigational situations with respect to the
surroundingsofanindividualship.Itis
notpossible
todeterminethenavigationalsafetyzone.
Hasegawa (1997) proposed calculating collision
risks(CR)byusingfuzzylogicintheriskassessment
of navigational areas. This model relies on the
calculated risk quantification that considers the
distancetotheclosestpointofapproach(DCPA)and
the time to the
closest point of approach (TCPA). It
also considers the decision‐making processes in
avoiding ship collisions. The limitation of this
approachisthattheindicatedriskonlyappliestothe
localvicinityofanindividualship.
Hara(1995)proposedasubjectivejudgmentvalue
(SJ) model to evaluate the ship route area
based on
navigationofficers’perception.Thismodelconsiders
factorssuchasthedistancebetweenships,theratesof
changeoftheships’directions,andtheirapproach.In
thismodel,therisksassociatedwiththefactorscanbe
quantified by reasoning rules derived from fuzzy
membership functions. The value of the
reasoning
rules was analyzed using a ship handling simulator
withnavigatorsservingasexperts.Inoue(1997,2000)
has proposed an environment stress (ES) model to
evaluate risk of a ship route based on a navigation
officer’s perception while operating a ship. The risk
wascalculatedbymeasuringthephysicalstresson
a
navigator and using a questionnaire. This risk
quantification model considers factors such as the
distance between a ship and another ship or an
obstacle,therate ofchangeoftherelativedirections,
and the approaching speed. It is a useful tool to
estimate the risk associated with a navigational
situation by assessing the navigation officer’s
difficultywithnavigatingaship.Asoneofthemajor
sources of human error, navigation officers play an
important role in navigating ships. However, this
type of approach can only be used to evaluate the
navigational safety of the surroundings of an
individualship.
In
previousstudies,thesemethodshaveshownto
beusefulforevaluatingtheriskassociatedwithship
navigation.However,twoadditionalpointsregarding
the safety evaluation model associated with ship
navigational situations have been considered in our
model.Thefirstoneisthatthequantificationofrisk
reflects the navigation officers’
perception for
estimatingrisksbetweenships.Atthesametime,the
evaluationofriskconsidersvariousadditionalfactors
in managing ship navigation safety in order to
support port authorities or vessel traffic service
centers. The second is that an algorithm has been
developedtoevaluatesafetyforanentireshiproute
area in real‐time. It aims to determine navigational
safetyzoneanywherealonganentireshiproutearea
ataspecifictime.Theaimofthisstudyistointroduce
a new model for estimating risk in an entire ship
route area in real‐time, which reflects navigation
officer’sperception.
2 ANEWSAFETYEVALUATINMODEL
A new safety evaluation model is presented in this
section for the evaluation of the safety in an entire
ship route area in order to support a port safety
authority or vessel traffic service center. This model
takes into consideration the navigation officer’s
perception while
navigating a ship in addition to a
varietyoffactors.Riskquantificationisincorporated
in this model, and a new algorithm for evaluating
safetyinanentireshiprouteareaisdeveloped.
2.1 Factorsasaffectinganavigationofficer’sperception
A safety index is developed to quantify risks that
reveal
the perception of a navigation officer
dependingonachangeinthenavigationalsituation.
It considers the process of an officer’s decision
making when encountering other ships as shown in
figure1.
Figure1. The process of a navigation officer’s decision
makingwhenencounteringotherships
A navigation officer takes action, whether giving
way or standing by, after recognizing the risks in a
given situation. Decision procedures for avoiding
risksarecomposedofthefollowingsteps.Atthefirst
step,dataiscollectedtoassesstheriskspresentedby
other ships. The risks of the encounters are
defined
based on the difference of direction, distance, and
speed. The second step considers the rules and a
ship’s maneuverabilityfor taking proper action. The
factors are designed considering the navigation
officer’s decision‐making process. The model
incorporates various factors that affect a navigator’s
perceptions during ship navigation. Factors are
classifiedaccordingtoshiprelatedinformation(ship
type, length of ship), relationship between the ships
(relative speed, distance between ships, encounter
situations) and environmental situations (time, day).
The detailed elements of each factor are shown in
Table1.
95
Table1.Thedesignoffactorsinsafetyindexmodel
_______________________________________________
ItemsDetails
_______________________________________________
Typeofship Containership,LNG,VLCC,Ferry,
Passengership,Bulkcarrier,Fisher,
LPG,PCC,Reefership,Tugboat
Lengthofship Under100m,101–150m,151–200m,201–
250m,251–300m,over301m
Relativespeed 0–1.0k’t,1.1–2.0k’t,2.1–3.0k’t,3.1–4.0k’t,
over4.1
k’t–
DistanceUnder5L,6–10L,11–15L,16–20L,21–
(L,lengthofship) 30L,over31L
Encounter Head ‐on Onthecenterlineoftheship
situations (give ‐way)showingfromrightahead
of30degreesabaftthebeamof
eithersideofship
Crossing Onthestarboardsideshowing

onfromrightaheadof
starboard 30degreesto112.5degrees
(give‐way)
Crossing Ontheportsideshowing
onport fromportaheadof
(stand‐on) 30degreesto247.5degrees
OvertakingAtthesternshowing67.5
(stand‐on) degreesfromrightaftoneach
side
ofship,
Time1stofficer’s04:00–08:00,16:00–20:00
(LT,localtime) 2ndofficer’s 00:00–04:00,12:00–16:00
3rdofficer’s 08:00–12:00,20:00–24:00
DayMon.,Tue.,Wed.,Thurs.,Fri.,Sat.,Sun.
_______________________________________________
2.2 Safetyindexforidentifyingrisksassociatedwith
navigationsituation
Thissectiondescribeshowtoidentifytheriskofeach
factorinthismodel.Aquestionnaireisusefultoolto
measure the degree of risk. In the questionnaire,
navigationofficerswereaskedhowmucheachfactor
affectstheirperception,using
anine‐levelevaluation
scale (level 1: no influence; level 9: significant
influence).Theresultsreflectthenavigators’opinions
inaquantitativemannerthatcanbeincorporatedinto
the safety evaluation model. In this model, each
elementinquestionisquantifiedusingequation(1):
1
1
N
ij ij
IR
N

 (1)
where:
Iij‐average of numerical values for j
th
element of i
th
item
Rij‐answervalueforj
th
elementofi
th
item( =1‐7)
N‐numberofrespondents
i‐itemnumberofquestionnaire(i=1‐8)
j‐elementnumberofeachitem
The results of the quantification of each factor
obtainedusingthequestionnaireisshowninTable2.
The safety level reflecting a navigator’s perceptions
canbecalculatedusingthesefactors.
Table2.Riskquantificationofeachfactordeterminedusing
questionnaire
_______________________________________________
Items Score
_______________________________________________
Typeofship 5.3‐8.1
Lengthofship 4.5‐8.1
Relativespeed 5.1‐7.5
Distance(L,lengthofship) 3.8‐7.8
Encounter Head‐onPassing Meeting
situations (give‐way) 7.9 4.0
Crossingon starboard 7.9 3.2
(give‐way)
Crossingon port 7.3 2.2
(stand‐on)
Overtaking 7.4 2.4
(stand‐on)
Time(LT,Localtime) 4.38‐5.50
Day 4.91‐5.08
_______________________________________________
2.3 Procedureforevaluatingsafetyinanentireshiproute
Figure2shows thestepwiseprocessofthealgorithm
toevaluatethesafetythroughoutanentireshiproute
area.Firstly,thewholeshiprouteareaisdividedinto
small sections and then ship data are collected for
each section. All
the ships in each section are
regarded as individual ships. In each section, target
ship data are collected within a specified range of
each ship. The collected ship data are defined as
factors in safety index, which includes ship
information, the relationship between ships and the
environmentalsituation(Figure3).
Figure2.Theprocessofevaluatingsafetyofanentireship
routearea
96
Figure3. Procedure for calculating safety level in each
section
Thelevelofsafetyineachsectioniscalculatedby
summing the quantified risks associated with the
factorsinthismodelusingequation(2):
11
ni
ij
SI I

 (2)
where:
SI‐Safetyindexforeachsection
n‐Numberofshipsineachsection
I
ij‐Riskquantificationofeachelementinquestion
i‐itemnumberofquestionnaire(i=1‐7)
j‐elementnumberofeachitem
Usingtheproposedalgorithm,thesafetylevelcan
be calculated for an entire navigational area, which
reflects the navigation officer’s perception. As a
result, a representative value for each section is
assignedasafetyindex.
2.4 SimulationMethod
In this simulation, automatic identification system
(AIS)
data have been used to reproduce a marine
traffic situation and to evaluate the navigation
situation,asshowninFigure4.
Figure4. Marine traffic simulation and safety index for
evaluatingriskofanavigationsituation
TheAISisanavigationdevicethattransmitsship
informationdataautomatically.TheAISisequipped
for domestic ships of over 500 GT and international
ships of over 300 GT by the International Maritime
Organization(IMO).AISdataareclassifiedaseither
static or dynamic, and consist of maritime mobile
service
identity (MMSI) number, ship name, current
shipposition, speed overground, and trueheading,
along with other variables. These data make it easy
for the analysis of both the whole traffic flow and
individualshipmovement.
3 SIMULATIONRESULTS
3.1 Subjectobservationarea
Asimulationwascarriedouttovalidate
theproposed
model for use as a safety evaluation model. It was
conducted for Osaka Bay, as shown in Figure 4.
OsakabayisJapan’slargestsemi‐enclosedsea,which
islocatedattheeasternendofSetoInland.Thisbay
has two entrances for the Osaka/Kobe port areas,
which are
the Akashi Strait and the Tomogashima
Channel. According to the Port Authority of Japan
(2010), the area used is latitude N34°14´ to N34°46´
and longitude E134°54´ to E135°26´. This simulation
was carried out using AIS data taken from the AIS
receiverinKobeUniversity.
Figure5.TrafficrulesinOsakaBay(Fromportauthorityin
Japan,2010)
Figure6showsthetrajectoriesofshipsnavigating
inOsakaBaybasedonAISdatabeginningat17:00for
a duration of one hour. In order to evaluate risk in
OsakaBay,thewholeshiprouteareaisdividedintoa
32x32mesh,withthesizeofeachgridsectionequal
to
1 square mile. At a specific time, ship data are
collectedusingAISdata.Thecollecteddataarethen
distributed to each section. For calculating the
relationshipriskbetween ships,data fromeach ship
are collected from within a range of 3 miles. The
resultsareshowninFigures
7and8 asanexample.
Figure 7 shows which area forms a relatively high
trafficdensity area at19080sec.Theaveragenumber
ofshipencounterswithina3‐mileradiusisshownin
Figure 8. The averagenumber of ship encounters at
19080secisabout3ships.Based
oncollecteddata,the
safetyindexforeachsectioniscalculated.Inthenext
part, the result of safety evaluation using the
proposedmodelisdescribed.
97
Figure6.TrajectoriesofshippassingOsakaBayfrom05:00
to06:00onMarch1,2013
Figure7.Thenumberofships ineachsectionat19080sec
Figure8. The average number of ships encounters with
respecttoanindividualshipineachsectionat19080sec
3.2 ResultsofriskevaluationinOsakaBayusingsafety
indexmodel
Thispartshowstheriskcalculationresultsusingthe
proposed safety index model. These results are
plottedincolorwithrespecttothelevelofthesafety
index. Figure 9 shows the results using the safety
indexinthe
entireshiprouteinOsaka Bayat19080
sec.TheresultsaftertwominutesareshowninFigure
10. These results are called a hazard map in this
study. A comparison of Figures 9 and 10 shows the
safetyindex changes dependingon ship movements
inreal‐time.Itillustrates
thatthenavigationalsafety
zoneandhazardzone canbedetermined easilyand
quickly.
The safety index level at a specific section for a
durationof1 hourbeginningat15:00ispresentedin
Figures 11 and 12, respectively. In addition, it
describes changes depending on the speed and
numberof
shipsinthesefigures.Figure11showsthe
safetyindexobservedatsection27x23.Theaverage
safetyindexinthissectionindicatesthatitisaround
93.2. It shows that the lowest and highest level of
safetyindexis56.18and145.97,respectively.
Figure9.Hazardmapaccordingtolevelofthesafetyindex
at19080sec
Figure10.Hazardmapaccordingtolevelofthesafetyindex
at19200sec
Thesafetyindexisaffectedbythechangesinboth
the number of other ships encountered and their
speeds, as shown in Figure 11 as (a), (b) and (c). In
theseresultsshowninFigure11,thenumberofships
andspeedareininverselyproportionaltoeachother.
Onaverage,
forone‐hourdurationandwithrespectto
an individual ship navigating through the section,
there is one passing ship and abouttwo otherships
beingencountered.Figure12showsthesafety index
observedinsection16x21.Theaveragesafetyindex
in this section is approximately 91.2 as
shown in
figure12(a).Itshowsthatthelowestandhighestlevel
of safety index is 55.14 and 125.44, respectively. In
98
figure 12(b), there are 2 ships passing through this
sectionon average. Itshows that the heaviest traffic
occursat20100 sec,whilethe highestlevel ofsafety
indexoccursat20220sec.Inthiscase,thespeedisnot
affected by the number of ship as shown in figure
12(c).
Figure11.Theresultsforevaluatingnavigationsituationof
section 27 x 23 using safety index (a) safety index (b)
Number of ships encountering other ships (c) Average
speedofshipspassingthroughthissection
Figure12.Theresultsforevaluatingnavigationsituationof
section 16 x 21 using safety index (a) safety index (b)
Number of ships encountering other ships (c) Average
speedofshipspassingthroughthissection
4 CONCLUSIONS
Thispaperintroducesanewsafetyevaluationmodel
thatcanbeusedtosupportaportsafetyauthorityor
vessel traffic service center. This model takes into
consideration the evaluation of an entire ship route
area at specific time. The risk calculation results
reflectanavigationofficer’sperceptionusing
asafety
index. This index was established using a
questionnairetoobtaininputfromnavigationofficers.
The algorithm was proposed to calculate the risk
throughout whole ship route area. This new safety
evaluationmodelisproposedasamethodtoestimate
riskthroughoutanentireshiprouteareainreal‐
time.
In order to verify the usefulness of the proposed
model,ariskevaluationwasimplementedforOsaka
Bay.Thesafetyindexineachsectionisillustratedin
color according to level of risk, which is called a
hazard map in this study. This approach allows for
visualizationoftherisk.
Thismodelisexpectedtobe
able to serve as a new tool for determining hazard
zones more quickly and easily than is currently
possible with other navigation safety evaluation
methods through the centralized management of an
entire ship route in real‐time,. The procedure
developed in this study can
be used for supporting
the vessel traffic service centers and port safety
authorities.
REFERENCES
FujiiY.,ShiobaraR. 1971,Theanalysisof traffic accidents,
TheJournalofNavigation,24(4):537‐543.
MacduffT.1974,Theprobabilityofvesselcollisions,Ocean
Industry:144‐148.
Goodwin E.M. 1975, A statistical study of ship domains,
TheJournalofNavigation,28:329‐341.
Fujii Y., Tanaka K. 1971, Traffic
capacity, The Journal of
Navigation,24:543‐552.
PietrzykowskiZ.2008,Ship’sfuzzydomain‐acriterionfor
navigational safety in narrow fairways, The Journal of
Navigation,61:499‐514.
Pietrzykowski Z., Uriasz J. 2009, The ship domain – a
criterion of navigational safety assessment in an open
seaarea,The
JournalofNavigation,62:93‐108.
WangN.2010,Aninterlligentspatialcollisionriskbasedon
thequaternionshipdomain,TheJournalofNavigation,
63:733‐749.
Hasegawa K., et al. 1997, Reconfiguration of ship auto‐
navigation fuzzy expert system(SAFES), The Kansai
SocietyofNavalArchitectsofJapan,8:191‐196.
Hara K., Nakamura S. 1995, A comprehensive assessment
system for the maritime traffic environment, Safety
Science,19:203‐215.
InoueK,KubonoM.,MiyasakaM.,HaraD.1997,Modeling
ofmariners’perceptionofsafetywhenbeingfacedwith
imminent danger, The Kansai Society of Naval
ArchitectsofJapan,97:235‐
245.
Inoue K. 2000, Evaluation method of ship handling
difficulty for navigation in restricted and congested
waterways,TheJournalofRoyalInstituteofNavigation,
53:167‐180.
The Osaka harbor information center for security of ship
navigation2010,PortofOsakaEntranceandDeparture
Manual