COURSECODE:

OE781CE
COURSETI
TLE:
ROADSAFETY
MEASURES
OE801CE
ROADSAFETYENGI
NEERI
NG
(
OPENELECTI
VE)
I
nst
ruct
ion: 3Per
iodsperweek
Dur
ati
onofUni
ver
sit
yExami
nat
ion: 3Hour
s
SUMMATI
VEEXTERNALEXAMI
NATI
ON:
70Mar
ks
CUMULATI
VEI
NTERNALEXAMI
NATI
ON:
30Mar
ks

WEEK PERIODS TOPI

C REMARKS
PER
WEEK

1 1 I
ntr
oducti
ont
othesubjectRoadsaf
ety 1
measur
esandimpor
tanceofthesubject

2 UNI
TI‐
Roadacci
dent
s 2

Causes,sci ent
ifi
c inv
esti
gations and data
col
lecti
on,analysi
sofi ndi
vidualacci
dentsto
arr
iveatr ealcauses,stat
isti
calmet hodsof
analy
sisofaccidentdat
a

3 Basic concepts ofroad acci

dentstat
ist
ics, 3
saf
ety performance functi
on :
The empir
ical
bayesmet hodi dent
if
icat
ionofhazardsroad
l
ocation

2 1 Appl
icat
ionsofcomput
eranal
ysi
sofcomput
er 4
dat
a.AssignmentI

2 UNI
TII‐
Saf
etyi
nRoaddesi
gn 5

Oper
ati
ngtheroadnet
workforsaf
ety
,hi
ghway
oper
ati
onandcount
ermeasures

3 Road saf
etyaudit
,pri
ncipl
es‐pr
ocedur
esand 6
pr
acti
ce,codeofgoodpract
iceandcheckl
i
sts

3 1 Road saf
etyaudit
,pri
ncipl
es‐pr
ocedur
esand 7
pr
acti
ce,codeofgoodpract
iceandcheckl
i
sts

2 Vehi
cledesi
gnfact
ors&Dr
iverchar
act
eri
sti
cs 8
i
nfl
uenci
ngroadsaf
ety

3 Di
scussi
on,
Assi
gnmentI
I 9
 4 1 UNI
TII
I‐Roadsi
gnsandTr
aff
icsi
gnal
s 10

Cl
assi
fi
cat
ion,locat
ionofsi
gns,measur
esof
si
gneff
ect
iveness

2 Typesofvisual
per
cept
ion,
signr
egul
ati
ons, 11
si
gnv i
sibi
l
ity

3 Si
gnv
ari
abl
es,
textv
ersussy
mbol
s 12

5 1 Roadmarki
ng:Roleofroadmar
kings, 13
cl
assi
fi
cati
on,
visi
bil
i
ty

2 Traf
fi
csignal
s:Need,si
gnal
face,
Il
lumi
nat
ion 14
andlocat
ionofsi
gnal
s,

3 Desi
gnofsi
gnal
s,Ar
eaTr
aff
icCont
rol 15

WEEK PERIODS TOPI

C REMARKS
PER
WEEK

6 1 Deli
neat
ers,Traf
fi
cImpactAttenuat
ors,Road 16
si
de restareas,Saf
ety bar
ri
ers,Traff
ic ai
d
posts

2 UNI
TIV 17

Traff
ic managementtechni
ques :Int
egr
ated
safet
y i mprovement and traf
fi
c calming
schemes,sppedandloadli
mit

3 Traf
fi
cli
ghts,saf
etycamer
as,t
est
sondr
iver 18
andvehi
cles

7 1 Pedest
ri
ansaf
etyi
ssues,
par
kingenf
orcement 19

2 Parki
ng enf
orcementi
nfl
uenceon acci
dent
s, 20
tr
aveldemandmanagement

3 Met hods oftr

aff
ic managementmeasures: 21
restri
cti
on oftur
ning movement
s,one way
streets
8 1 Ti
dalfl
ow operat
ion methods,excl
usi
vebus 22
l
anesandcl
osingside‐
str
eets

2 Lat
esttool
s and t
echni
ques used f
orr
oad 23
saf
etyandt
raf
fi
cmanagement

3 Roadsafetyi
ssuesandv ari
ousmeasuresf
or 24
road saf et
y, Legi sl
ation, Enf
orcement,
Educat
ionandpropaganda

9 1 Airqual
it
y,Noi
seandEner
gyI
mpact
s;Costof 25
roadacci
dent
s

2 UNI
TV 26

AccidentManagement:
I
ntroducti
on,Char
acter
ist
icsoft
raf
fi
c
i
ncidents,
typesofinci
dents

3 I
mpacts, Incident management pr
ocess, 27
I
nci
denttr
affi
cmanagement

10 1 Appl
icat
ions of I
TS: Motori
st inf
ormati
on, 28
Equi
pmentused;Pl anni
ng ef
fect
iveinci
dent
managementprogr
am

2 Best pr
act
ice i
n I
nci
dent management 29
progr
ams

3 National i mpor tance of sur vi

v al of 30
Transportati
on systems dur
ing and afterall
natural di sast
ers especi al
l
y cy clones
,earthquakes,fl
oodsetc.

11 1 National i mpor tance of sur vi

v al of 31
Transportati
on systems dur
ing and afterall
natural di sast
ers especi al
l
y cy clones
,earthquakes,fl
oodsetc.

2 National i mpor tance of sur vi

v al of 32
Transportati
on systems dur
ing and afterall
natural di sast
ers especi al
l
y cy clones
,earthquakes,fl
oodsetc.

Nati
onalImpor
tanceofsurvi
v ali
nman-
made 33
di
saster
sli
kesabotage,
ter
ror
ism et
c.
3
Cour
seCode:
OE801CE

Cour
set
it
le:
RoadSaf
etyEngi
neer
ing

Facul
tyName:
Mrs.Shwet
aKaushi
k

Cour
seObj
ect
ives:

Int
roduct
iontothewaysofcol
lect
ionofAcci
dentdat
aanduseof
computersforanal
ysi
sofacci
dentdata

I
ntr
oduct
iont
ovar
iousf
act
orsconsi
der
edf
orr
oadsaf
etyand
management

 Expl
aint
her
oadsaf
etyappur
tenancesanddesi
gnel
ement
s

 Di
scusst
hev
ari
oust
raf
fi
cmanagementt
echni
ques

 Explai
nthetechni
quesofmanagementbef
oreandaf
tert
he
i
ncidenthasoccurr
ed
UNI
TI

ROADACCI
DENTS

Causes,Scienti
fi
ci nvesti
gati
ons and dat a col
lect
ion,Analysis of
i
ndivi
dualaccident
st o arri
ve atrealcauses,stati
sti
calmet hods of
analy
sisofaccidentdata,BasicconceptsofRoadacci dentst
atisti
cs,
Safet
yperf
ormancef uncti
on:Theempi r
icalBayesmethodI dent
if
ication
ofHazardsroadlocati
on.Applicat
ionofcomput eranalysi
sofacci dent
Data

Ov
erv
iew

Thislect
urecov er
soneoft hemosti mpor t
antnegativ
ei mpactsof
tr
ansport
ati
onsy st
em,namel yt
heacci dent
s.Thislect
urefir
stpresents
some intr
oductory st
uffincl
uding some sal i
entaccidentstati
sti
cs,
causesofaccidents,acci
dentdatacollect
ion,acci
dentreconst
ructi
on,
safet
ymeasuresandsaf et
yaudit
.

Saf
etyPer
for
manceFunct
ion

SPFsar ecr ashpr edicti

onmodel s.Theyar eessentiall
ymat hematical
equat i
onst hatrelatet henumberofcr ashesofdi fferenttypest osite
charact erist
ics.Thesemodel salway sincludet r
affi
cv olume( AADT)but
mayal soi ncludesitechar acter
isti
cssuchasl anewidth, shoul
derwidth,
radius/degr ee of hor izontal curves,pr esence of t urn lanes (at
i
nter sections),andt raffi
ccont r
ol( atintersecti
ons).Oneexampl ei
sthe
foll
owi ngSPFf rom Saf etyAnal ystf orpr edi
ctingthet ot
alnumberof
crashesonr ural
mul til
anedi vi
dedr oads:
PLe5.05(AADT)
0.
66

Wher
e:

Pi
sthet
otal
numberofcr
ashesi
n1y
earonasegmentofl
engt
hL.
Thepr i
marypurposeoft hisSPFf rom Saf et
yAnal ystistoassi stan
agencyinthei
rnetworkscr eeningpr ocess,i
.e.
,toidentif
ysit
est hatmay
benefi
tfr
om asafetytr
eatment .Thisisar el
ati
velysi
mpl eSPFwher et
he
predi
ctednumberofcrashespermi leisaf unct
ionofj ustAADT.Ont he
otherhand,Bauerand Har wood ( 2012)pr ovide a mor e complex
predi
cti
onmodel f
orfatalandi njurycrashesonr uralt
wol aneroads:

Wher
e:

NFI = f
atal
-and-
inj
urycr
ashes/
mi/
yr

AADT = v
eh/
day

G = absolut
ev al
ueofpercentgr
ade;0per
centf
orl
evel
tangent
s;≥1percent

Ot
her
wise

R = cur
ver
adi
us(
ft)
;mi
ssi
ngf
ort
angent
s

I
HC = hori
zontalcur
vei
ndi
cat
or:1f
orhor
izont
alcur
ves;0
other
wise

LC = horizont
alcur
vel
engt
h (
mi)
;not appl
i
cabl
efor
tangents

l
n = nat
ural
logar
it
hm f
unct
ion

b0,
…,b4 = r
egr
essi
oncoef
fi
cient
s

Thi
spr edict
ionmodelwasest
imatedusi
ngdatafrom Washi
ngt
on.The
goaloft hi
smodelwastoexaminethesaf
etyaspect
sofhori
zont
aland
ver
ti
cal curvat
ure.
I
ntr
oduct
ion

Thepr oblem ofacci dentisav er yacut einhighwayt ranspor t

ati
onduet o
compl exflowpat ternofv ehicul artraffic,
presenceofmi xedt r
affi
calong
withpedest ri
ans.Traf fi
cacci dentl eadst olossofl if
eandpr opert
y.Thus
thetr aff
icengi neershav etounder takeabi gr esponsibili
tyofprovidi
ng
safet raff
icmov ement stot her oaduser sandensur etheirsafety.Road
accident scannotbet otal
lypr ev entedbutbysui tabl
et raffi
cengineeri
ng
andmanagementt heacci dentr atecanber educedt oacer t
ainextent.
Fort hisreasonsy stemat icst udyoft raffi
cacci dentsar erequir
edtobe
carri
edout .Properi nvesti
gat ionoft hecauseofacci dentwi l
lhelpto
proposepr eventi
vemeasur esi nt er
msofdesi gnandcont r
ol.

Obj
ect
ivesofacci
dentst
udi
es

Someobj
ect
ivesofacci
dentst
udi
esar
eli
stedbel
ow:

 To study the causes of acci

dent
s and suggest cor
rect
ive
measuresatpot
entiall
ocat
ion

 Toev
aluat
eexi
sti
ngdesi
gn

 Tocomput
ethef
inanci
all
ossesi
ncur
red

 To supportthe pr
oposed desi
gn and prov
ide economi
c
j
ust
if
icat
iont
othei
mprov
ementsuggest
edbyt
hetr
aff
icengi
neer

 To carr
youtbefor
eandafterst
udi
esandt
o demonst
rat
ethe
i
mpr ov
ementi
ntheprobl
em.

Causesofr
oadacci
dent
s
 Thev
ari
ouscausesofr
oadacci
dent
sar
e:

 RoadUser s-Excessi
vespeedandr ashdri
ving,vi
olat
ionoftr
aff
ic
rul
es,fail
ureto percei
vet r
affi
c si
tuat
ion orsign orsignalin
adequateti
me,carel
essness,f
ati
gue,al
cohol
, sl
eepetc.

 Vehicl
e-Defect
ssuchasf
ail
ureofbr
akes,st
eer
ingsy
stem,t
yre
burst,
li
ght
ingsy
stem .

 RoadCondi
ti
on-Ski
ddi
ngr
oadsur
face,
pothol
es,
rut
s.

 Roaddesign-Def ecti
vegeomet r
icdesignli
kei nadequat
esight
di
stance,i
nadequatewidthofshoulders,i
mpr opercurvedesi
gn,
i
mpr opert
raf
fi
ccontroldev
icesandimproperl
ighti
ng,.

 Envi
ronment
alfact
ors- unf
avorabl
eweathercondit
ionsli
kemist
,
snow,smokeandheav yrai
nfal
lwhichr
estr
ictnor
malv i
si
bil
it
yand
andmakesdriv
ingunsafe.

 Othercauses-
improperl
ocat
ionofadver
ti
sementboar
ds,gat
eof
l
evelcrossi
ngnotcl
osedwhenrequi
redet
c..

Acci
dentst
ati
sti
cs

Thest ati
sticalanaly sisofacci denti scar ri
edoutper iodical l
yatcr itical
l
ocat ionsorr oadst r etcheswhi chwi llhel
pt oar riveatsui t
abl emeasur es
toef f
ect i
velydecr easeacci dentr ates.Itisthemeasur e( orest imat es)of
thenumberandsev erit
yofacci dent .Thesest at i
sti
csr epor tsaret obe
mai ntainedzone- wi se.Acci dentpr onest retchesofdi fferentr oadsmay
beassessedbyf indi ngtheacci dentdensi typerl engthoft her oad.The
placesofacci dentsar emar kedont hemapandt hepoi ntsoft heir
clustering(BLACKSPOT)ar edet er mined.Byst at isti
cal studyofacci dent
occur renceatapar t i
cularroadorl ocationorzoneofst udyf oral ong
periodoft imei tispossi blet opr edictwi thr easonabl eaccur acyt he
probabi li
tyofacci dentoccur renceperdayorr el ativ
esaf etyofdi fferent
classesofr oaduseri nthatlocat ion.Thei nterpr etati
onoft hest ati
st ical
datai sv eryimpor tantt opr ovidei nsightt othepr oblem.Theposi ti
onof
Indiai nt hey ear2009 i n count ry-wisenumberofper son ki l
led per
100000populati
onsasshowni ntheFi
gure1andtheincr
easeinrat
eof
acci
dentfrom year2005toyear2009isshowninthetabl
e.1.In2009,
14accidentsoccur
redperhour
.

Figure1:Count
ry-
wisenumberofpersonkil
l
edper100000populat
ions
(Ref.Mini
str
yofRoadTranspor
tandHighwaysTr
anspor
tResear
chWi ng)

Tabl
e1:NumberofAcci
dent
sandNumberofPersonsI
nvolved:2001to
2009(
Ref
.Mini
str
yofRoadTranspor
tandHi
ghwaysTransportResear
ch
Wing)
Acci
dentAnal
ysi
s

Acci
dentdat
acol
l
ect
ion
Theaccidentdatacol
lect
ionisthefir
ststepint heacci
dentstudy
.The
datacoll
ecti
onoftheaccident
sisprimari
lydonebyt hepoli
ce.Motori
st
accidentrepor
ts ar
e secondary data which arefil
ed by motori
sts
themselves.Thedatat o becoll
ected should compri
sealloft hese
parameters:

1.General-Dat
e,ti
me,personi
nvol
vedi
nacci
dent
,cl
assi
fi
cat
ionof
acci
dentli
kefat
al,
ser
ious,mi
nor

2.Locat
ion-Descr
ipt
ionanddet
ail
ofl
ocat
ionofacci
dent

3.Detail
sofv ehicl
einvol
ved-Registrat
ionnumber
,descr
ipt
ionof
vehicl
e,l
oadingdetai
l
,vehi
cul
ardefects

4.Nature ofacci
dent-Det
ail
s ofcol
l
isi
on,damages,i
njur
yand
casual
ty

5.Road and t
raff
iccondit
ion-Det ai
lsofroad geomet
ry,sur
face
char
acter
ist
ics,
typeoft
raff
ic,
traf
fi
cdensi
tyet
c..

6.Pri
mar ycausesofaccident-Det
ail
sofv ari
ouspossibl
ecases
(al
readymentioned)whi
char
ethemaincausesofacci
dent
.

7.Accidentcost-Fi
nanci
all
ossesi
ncur
redduet
opr
oper
tydamage,
personali
njur
yandcasual
ty

Thesedatacoll
ectedneedpr
operst
ori
ngandr
etr
iev
ingf
ort
hef
oll
owi
ng
purpose.Thepur
posesareasfol
l
ows:

1.I
denti
fi
cati
onoflocat
ionofpoi
ntsatwhi
chunusual
l
yhi
ghnumber
ofacci
dentoccur
.

2.Det
ail
ed funct
ionalev
aluati
on ofcr
it
icalacci
dentl
ocat
ion t
o
i
denti
fyt
hecausesofaccident
s.

3.Devel
opmentofprocedurethatall
owsident
if
icat
ionofhazar
ds
bef
orelar
genumberofacci
dentsoccur
s.

4.Developmentofdif
ferentstati
sti
calmeasuresofv ar
iousacci
dent
rel
atedfact
orstogivei nsi
ghti
ntogeneralt
rends,commoncasual
fact
ors,dr
iverpr
ofi
l
es, etc.

Acci
denti
nvest
igat
ion
The acci
dentdat
a col
l
ecti
on inv
olv
es ext
ensi
vei
nvest
igat
ion whi
ch
i
nvol
vesthefol
l
owingpr
ocedur
e:

1.Repor
ti
ng:
Iti
nvol
vesbasi
cdat
acol
l
ect
ioni
nfor
m oft
womet
hods:

 Motori
stacci
dentrepor
t-I ti
sf i
led byt
he i
nvol
ved mot
ori
st
i
nvolv
edinall
acci
dentsf
atal
ori
njur
ious.

 Policeaccidentrepor
t-Iti
sfil
edbyt heattendantpoli
ceoffi
cerfor
allaccidentsatwhichanoffi
cerispresent
.Thi sgeneral
lyi
ncludes
fatalacci dents or most
ly accidentsinv ol
ving seri
ous inj
ury
requiredemer gencyorhospi
taltreatmentorwhi chhav ei
ncurred
heav ypropertydamage.

2.AtScene-Investi
gati
on:I
tinv
olvesobtaininginformationatscene
suchasmeasur ementofskidmar ks,exami nati
onofdamageof
vehi
cles,photographoffi
nalpositi
onofv ehi
cles,examinati
onof
condi
tionandf uncti
oni
ngoftraff
iccontroldevicesandot herroad
equi
pment s.

3.TechnicalPreparat
ion:Thisdata col
lect
ion step i
sneeded for
organizat
ionandi nter
pret
ati
onoft hestudymade.I nthisst
ep
measur ementofgrades,si
ghtdist
ance,pr
eparingdrawingofaf
ter
accidentsit
uati
on,determi
nati
onofcr i
ti
calanddesi gnspeedfor
curvesisdone.

4.ProfessionalReconstructi
on:Int hi
s step ef f
ortis made t o
determinef r
om whateverdatai
savail
ablehowt heaccidentoccurs
fr
om t he av ai
l
able data.Thisinvol
ves acci
dentr econstruct
ion
whichhasbeendi scussed underSecti
onNo. 7indet ai
ls.Itis
professi
onallyr ef
err
ed as det er
mining behavior
alcauses of
accident.

5.CauseAnal ysi
s:I
tistheef
fortmadet
odeter
minewhytheacci
dent
occurred f
rom thedat aavail
abl
eand t
heanalysi
sofaccident
reconstr
ucti
onstudies.

Acci
dentdat
aanal
ysi
s

Thepurposei
st ofi
ndthepossibl
ecausesofacci
dentr
elat
edtodri
ver
,
vehi
cle,and roadway
. Accident anal
yses ar
e made t o dev
elop
i
nfor
mat
ionsuchas:

1.Dri
verandPedestr
ian-Accidentoccurr
encebyagegr oupsand
rel
ati
onshi
ps of acci
dents to phy si
cal capaci
ti
es and to
psychol
ogi
cal
testr
esul
ts.

2.Vehicl
e-Accidentoccur
rencerel
atedt
ochar act
eri
sti
cofv ehi
cl
e,
severi
ty,
locat
ionandext
entofdamagerel
atedtovehicl
es.

3.Roadwaycondit
ions-Rel at
ionshi
psofaccidentoccur
renceand
sever
it
ytocharacter
ist
icsoftheroadwayandr oadwaycondi
ti
on
andrel
ati
vev
aluesofchangesr el
atedt
oroadways.

Itisi mportantt o compute accidentr at

e whi ch ref
lectacci
dent
i
nvolvementby t ype ofhighway.These r ates provi
de a means of
comparingtherelat
ivesaf
etyofdif
ferenthi
ghwayandst r
eetsyst
em and
tr
aff
iccontrols.Anotheri
saccidentinvolv
ementbyt hetypeofdri
vers
andvehicl
esassociatedwit
haccidents.

Acci
dentr
econst
ruct
ion

Acci
dentreconst
ruct
iondealswit
hrepresenti
ngtheacci
dent
soccur r
ed
i
nschemat i
cdiagram todeter
minethepre-col
li
si
onspeedwhichhelps
i
nregul
ati
ngorenf or
cingrul
estocontr
olorcheckmovementofvehicl
es
onroadathighspeed.Thef ol
l
owingdataarerequi
redtodet
erminethe
pr
e-col
l
isi
onspeed:

Massoft
hev
ehi
cle

Vel
oci
ti
esaf
tercol
l
isi
on

Pat
hofeachv
ehi
cleasi
tappr
oachescol
l
isi
onpoi
nt

Below inFigure1aschemat i
cdi agr
am ofcol l
isi
onoft wov ehi
cl
esi s
shown t hatoccurdur ing t
urning movement s.This diagr
am is also
knownascol l
isi
ondi agram.Eachcol li
sionisrepresentedbyasetof
arr
ows t o show the di r
ecti
on ofbef ore and aftermov ement.The
col
lisi
on diagram pr ovides a power fulvisualr ecor
d of accident
occurrenceoverasignifi
cantperiodofti
me.
 Figure1: Col
l
isi
on
diagram of t
wovehi
cles

The coll
isi
onmay
beoft wo ty
pes
col
linear i
mpactor
angular coll
isi
on.
Beloweach ofthem ar
e
describedin detail
.
Colli
nearimpactcanbeagai
ndi
vi
dedi
ntot
wot
ypes:

1.Rearendcol
l
isi
on

2.Head-
oncol
l
isi
on.

I
tcanbedet
ermi
nedbyt
wot
heor
ies:

1.Poi
ssonI
mpactTheor
y

2.Ener
gyTheor
y

Poi
ssoni
mpactt
heor
y

Poi
ssonimpacttheor
y ,
divi
desthei
mpacti
ntwopar
ts-compressi
on
andrest
it
uti
on.TheFigure2showstwovehi
cl
est
ravel
l
ingatanini
ti
al

speedof and col l

idesandobtai
naunifor
m speedsay atthe
compressi
onst
age.Andaftert
hecompressi
onstageisovert
hef
inal

speedis and .Thecompr

essi
onphasei
sci
tedbyt
hedef
ormat
ion
ofthecar
s.
 Fi
gur
e2:
Compr
essi
onPhase

Fr
om Newt
on’
ssecondl
aw

(
1)

where, and aret

hemassesoft hecarsand i sthecontactforce.We
knowt hateveryreact
ionhasequalandopposi
teaction.Soasther earvehicl
e
pushest hevehicl
eaheadwithfor
ce .Thev ehi
cleaheadwi l
lalsopusht herear
vehi
clewi t
hsamemagni t
udeofforcebuthasdif
ferentdir
ect
ion.Theact i
onf or
ceis

represent
edby ,
wher
east
her
eact
ionf
orcei
srepr
esent
edby asshowni
n
Figure3.

Fi
gur
e3:
For
ceappl
i
edoneachv
ehi
cle

I
nt hecompressionphasecar sar
edef or
med.Thecompressi
onphaseter
minat
es
whent hecarshav eequalv
eloci
ty.Thusthecar
sobtai
nequalvel
oci
tywhi
ch
generatest
hefollowingequati
on:

(
2)

wher
e, whichist
hecompressi
oni mpulseand i st
he
compressi
ont
ime.Thus,t
hev el
oci
tyaf
tercol
li
sionisobtai
nedas:
 (
3)

Thecompr
essi
oni
mpul
sei
sgi
venby
:

(
4)

I
nther
est
it
uti
onphaset
heel
ast
icpar
tofi
nter
nal
ener
gyi
srel
eased:
I

(
5)

(
6)

Where, i
stheresti
tut
ioni
mpulseand i stherest
it
uti
onimpul
se
andistherest
it
uti
onti
me.Accordi
ngtopoisonshy
pothesi
srest
it
uti
onimpul
seis
propot
ionalt
ocompressi
onimpulse.

(
7)
Rest
it
uti
oni
mpul
se i
sgi
venby:

(
8)

Thet
otal
impul
sei
s

(
9)

Theposti
mpactv
eloci
ti
esar
egi
venby
:

(
10)

(
11)

Wher
e .Butwear
erequi
redt
odet
ermi
net
hepr
e–col
l
isi
onspeed

accor
dingt
owhi
cht
hesaf
etyont
her
oadcanbedesi
gned.Sowewi
l
ldet
ermi
ne
and

Fr
om t
hegi
venv
aluesof and

Ener
gyt
heor
y

Appl
ying princi
ple of conserv
ati
on of energy or conserv
ation of
moment um alsotheini
ti
alspeedofthevehicl
ecanbecomput edifthe
ski
dmar ksar eknown.I
tisbasedont heconceptt
hatthereisreduct
ion
i
nkineti
cener gywit
hthewor kdoneagai
nsttheski
dresist
ance.Soifthe

vehicleofweight sl owdownf r
om speed t o ,t hent
helossin
kineti
cenergywi l
lbeequaltot hewor kdoneagai
nstskidr
esi
stance,
wher eworkdoneisweightofthev ehi
clemult
ipl
i
edbytheski
ddi
stance
andt heski
dresist
ancecoef
fi
cient.

(
1)

wher
e, i
stheski
dresi
stancecoef
fi
cientand i
stheski
ddi
stance.I
t

al
sofol
l
owsthelawofconser
vati
onofmoment
um ( , ar et
he
massandv
eloci
tyoff
ir
stvehi
clecol
l
idi
ngwi
thanot
herv
ehi
cl
eofmass

andv
eloci
ty , r
espect
ivel
y)
(
2)
Angul
arcol
l
isi
on

Angularcollisi
on occurs when two v ehicl
es comi ng atri
ghtangl es
coll
ieswitheachot herandbif
urcatesi
ndi ffer
entdirect
ion.Thedirecti
on
ofthev ehi
clesaf t
ercoll
isi
oninthiscasedependsont heinit
ialspeeds
oft hetwov ehicl
esandt hei
rwei ghts.Onegener alcasei sthatt wo
vehicl
escomi ngf r
om southandwestdi recti
onaftercoll
idi
ngmov einit
s
result
antdir
ectionasshowni nFigure1.
 Fi
gur
e1:
Angul
arcol
l
isi
onoft wov ehicl
esresul
ti
ngi
nmov
ementi
n
resul
tantdirect
ion

Themassoft
hecar1i
s kgandt
hecar2i
s kgandt
hei
nit
ial

veloci
tyi
s m/ sand m/ sr espect
ivel
y.Soasthemoment
um i
sthe
productofmassandvel
oci
ty.Themoment um oft
hecar1andcar2is

kgm/sand kgm/srespectiv
ely
.Byt helaw ofconser
vati
on
ofmomentum the fi
nalmoment um should be equaltot he i
nit
ial
momentum.Butasthecarar eapproachi
ngeachot heratananglethe
fi
nalmomentum should notbe justmer e summat i
on ofbot hthe
momentum butthe resul
tantoft he two,Resultantmoment um =

kgm/
s.Theangl
eatwhi
cht
heyar
ebi
fur
cat
ed

aft
ercol
li
sioni
sgivenby wherehi
sthehypotenuseandbis
thebase.Ther
efor
e,thecar
sar
eincl
inedatanangle.Incl
inedatan

angle= .Now,sincethemassofthetwovehi
clesare
same the f
inalvel
oci
tywillpr
oporti
onall
ybe changed.The gener
al
schemati
cdiagr
amsofcol
lisi
onareshowninFigs.2t
o 4.
Fi
gur
e2:
Aft
ercol
l
isi
onmov
ementofcar1nor
thofwestandcar2i
n
eastofnor
th

Fi
gur
e3:
Aft
ercol
l
isi
onmov
ementofcar1andcar2i
nnor
thofeast
 Fi
gur
e4:
Aft
ercol
l
isi
onmov
ementofcar1nor
thofeastandcar2i
n
sout
hofeast

Saf
etymeasur
es

Theul ti
mat egoalistodevel
opcertai
nimprovementmeasur esto
miti
gat etheci r
cumstancesl
eadingtotheaccidents.Themeasur esto
decreaset heaccidentrat
esaregeneral
lydiv
idedintothreegroups
engineering,enfor
cementandeducation.Somesaf etymeasur esare
describedbel ow:

Saf
etymeasur
esr
elat
edt
oengi
neer
ing

Thevari
ousmeasur
esofengi
neer
ingt
hatmaybeusef
ult
opr
event
acci
dentsar
eenumerat
edbel
ow

Vi
sual
gui
dancet
odr
iver

Thereisconsecut
ivechangeofpict
urei
ndriver’smindwhileheisin
motion.Thenumberoff act
orst
hatthedri
vercandisti
ngui
shandcl ear
ly
fi
xinhismindisli
mi t
ed.Onanaveragethepercepti
ontimeforvi
sionis

,f
orheari
ngi
s andf ormuscularr
eacti
onis .The
numberoff
actor
sthatcanbet akenintoaccountbyorgansofsenseof
adri
veri
nonesecondisgiv
enbyt heformulabel
ow.
 (
1)

wher e, =No.off act

orst
hatcanbetakeni
ntoaccountbytheorgan
ofsenseofdr i
verf
or m long, =speedofv ehi
cl
einm/ sec.Fact
ors
affecti
ngdr
iversatt
ent
ionwhenheisonroadcanbedivi
dedintothr
ee
groups:

1.Factor
sr el
ati
ngtotheroad“elementsofroadthatdi
rect
lyaf
fect
thedri
vingofav ehi
clearet
raff
icsi
gns,changesindi
rect
ionof
road,t
hreeleggedint
ersect
ionandvari
ousotherthi
ngs.

2.Fact
orsconnect
edwi
tht
raf
fi
c“Ot
herv
ehi
cl
es,
cycl
es,
pedest
ri
ans.

3.Fact
orsr
elat
edi
ndi
rect
lyt
othev
ehi
cl
emot
ion.

4.“Bui
ldi
ngandst
ruct
urest
hatst
ri
ket
heey
e,v
eget
ati
on,
landscape,
etc.

Sousingthel
awsofv
isual
per
cept
ioncer
tai
nmeasur
eshav
ebeen
suggest
ed:

1.Contr
asti
nv isi
bil
it
yoftheroadshouldbeachievedbyprovi
sionof
el
ementsthatdiff
erfr
om it
ssur r
oundingbycolors,
pat
ternsuchas
shoul
derst
rips,shoul
dercoveredwit
hgr ass,
edgemarkings.

2.Pr
ovi
dingr
oadsi
dev
eget
ati
oni
sanef
fect
ivemeans.

3.Thevisi
bil
i
t yofcrownoftr
eesf
rom adi
stantl
ocat
ioni
sal
sov
ery
usef
ul i
nvisualguidi
ng.

4.Theprovi
sionofguardrai
l
sofdif
fer
entcont
rasti
ngcol
orsalso
takesdr
iversat
tent
ionandpr
eventf
rom monotonousdr
ivi
ng.

Figure1and2isavisual guidancemeasur
e.Pl
anti
ngtreesal
ongsideof
roadwaywhichhasat urningangleat
tr
actsat
tent
ionofthedr
iverand
signalst
hatatur
nispr esentahead.
 Fi
gur
e1:
Bif
urcat
ionoft
hehi
ghway

Thefigurebelow i
sanotherexample,whenthedir
ecti
onofroadhasa
hazardousat-
gradeinter
sect
iontr
eesareplantedi
nsuchawayt hatit
seemst hatt
hereisdensefor
estaheadanddriv
erautomati
cal
l
ytendst o
stoporreducethespeedofthevehicl
esothatnoconfl
ict
soccuratthat
point
.
 Fi
gur
e2:
Roadseemedt
obest
oppedbyadensef
orest

Drivertendstoextr
apolatet
hef urtherdi
rect
ionoftheroad.Soi
tist he
responsibil
i
tyofthetraff
icengineertomaket hedr
iverpsy
chol
ogicall
y
conf i
dentwhil
edriv
ingthatreducest heprobabil
i
tyoferr
orandprevent
ment alstr
ain.

Roadr
econst
ruct
ion

Thenumberofv ehiclesont heroadincreasesfrom yeartoy ear,which

i
ntroducescomplicationsintoor
ganizati
onoftraffi
c,sharpl
yreducest he
operati
onandt ranspor tat
ioncharacteri
sti
cofr oadsandl eadt ot he
growthofaccidentr ate.Thisl
eadstot heneedofr econstructingroad.
The places ofacci dents need to be properl
y mar ked so thatt he
reconst
ruct
ioncanbepl annedaccordi
ngly.

Fi
gur
e3:
Diagr
am ofacci
dent
sbef
oreandaf
terr
econst
ruct
ion

TheFi gure3showst hatt her

ewer et oomanyconf li
ctpointsbef ore
whichr educedt oaf ew numberaf t
erconstruct i
onofislandsatpr oper
places.Reconst ructi
onpr ocessmayal soincludeconstructi
onofanew
roadnextt ot heexisti
ngr oad,renewalofpav ementwi thoutchangi ng
the hor i
zontalalignmentorpr ofi
le oft he r oad,reconstruct
ion a
parti
cul arsectionofr oad.Few mor eexampl esofr econstr
ucti
onof
selectedr oadsectiontoimpr ovetraff
icsafetyar eshowninFigure4.
 Fi
gur
e4:
Roadr
econst
ruct
iont
echni
que

TheFi gur
e4( a)showssepar ati
onofdi recti
onofmai nst r
eam oft raf f
ic
from thesecondaryonesbyshi ft
ingplaceoft hree-l
eginter
sect i
on,
Figure4(b)showssepar ati
onofr oadswi thconstructi
onofconnect ion
betweent hem andFi gur
e4(c)showst heconst ructi
onofaddi t
ional lane
forturni
ngv ehicl
es.Theplussignindicatestheconf l
ictpoi
ntsbef ore
theroadr econstr
uctionhasbeencar ri
edout .Theaf t
erreconstruction
fi
gureshowst hatjustbyli
tt
leal
terati
onofasect ionofroadhowt he
confli
ctpointshavebeenr esol
vedandsmoot hflowoft hevehiclesinan
organizedmannerhav ebeenobtained.

Channel
i
zat
ion

TheChannel izati
onoft raffi
cati ntersect
ionseparat
esthet r
affi
cstream
trav
ellingi ndi f
ferentdirection,providi
ngthem aseparatelanethat
correspondst otheirconv enientpathandspr eadi
ngasf araspossible
thepoi ntsofconf li
ctbet weencr ossingtraff
icstr
eams.Thet raf
fi
cl anes
aresepar atedbymar kingr elevantli
nesorbyconst r
ucti
ngsl i
ghtl
y
elevatedi slandsasshowni nFigure5.ProperChanneli
zationreduces
confusi on.Thenumberofdeci si
onr equi
redtobemadebyt hedriverat
anyt i
mei sr educedal l
owi ngt hedr i
verti
met omakenextdeci si
on.The
pr
inci
plesofpr
operchannel
i
zedi
nter
sect
ionar
e:-

1.Thelayoutofi
nter
sect
ionshoul
dbev
isi
blycl
ear
,si
mpl
eand
underst
andabl
ebydriv
er.

2.Shoul
densur
esuper
ior
it
ytot
hev
ehi
cl
esusi
ngr
oadofhi
ghercl
ass.

3.Layoutofinter
sect
ionmakesitnecessar
yforadri
verrunni
ng
thr
oughittochooseateachmomentoft i
meoneofnotmor ethan
twopossibledir
ect
ionoft
ravel
.Thisisachiev
edbyvisual
gui
dance,isl
andsandmarkings.

4.Theisl
andprov
idedshoul
dsepar
atehi
ghspeed,
thr
oughand
tur
ningtr
aff
icf
lows.

5.Thewi dthoftr
aff
iclaneshouldensureunhamperedturni
ngtot
he
bigvehicl
es.Widt
hofst rai
ghtsecti
onwithoutker
bshouldbe3.
5
m andt hatoft
raf
ficlanenearisl
andis4.5-
5m atentryand6m at
exit
.

6.Pedest
ri
ancr
ossi
ngshoul
dbepr
ovi
ded

Fi
gur
e5:
Channel
i
zedI
nter
sect
ionensur
ingsaf
ety

Roadsi
gns

Roadsi
gnsar
eint
egr
alpar
tofsaf
etyast
heyensur
esaf
etyoft
hedr
iver
himself(war ningsi gns)andsaf etyoftheotherv ehiclesandpedest rians
onroad( r
egul atorysigns).Dr i
vershoul
dbeabl et or eadthesignf r
om a
dist
ancesot hathehasenought i
met ounder standandr espond.I tis
essentialthatt heyar einstal
ledandhav ecorrectshape, colour,sizeand
l
ocation.Itisr equiredtomai ntai
nthem aswel l,withoutmai ntenancei n
soundcondi ti
onj usttheirinstal
lmentwouldnotbebenef i
ci
al.Accor ding
toBrit
ishinv estigationheightoftextinroadsignshoul dbe

Where,N=No.ofwor
dsonthesign,v=speedofvehi
cl
e(kmph)
,L=
di
stancef
rom whi
chi
nscr
ipt
ionshoul
dbediscer
nibl
e(m)

Ot
hermet
hods

Var
iousot
hermet
hodsoft
raf
fi
cacci
dentmi
ti
gat
ionar
edescr
ibedbel
ow:

St
reetl
i
ght
ing

Streetl
ightningofappr opri
atestandardcont ri
butestosaf etyinurban
areaduringni ghtti
meduet opoorv i
sibi
li
ty.Instal
lat
ionofgoodl i
ghti
ng
result
sin21%r educti
oni nallacci
dents,29%r eductionin` `
allcasual
ty'
'
accidents,
21%r educti
oni n``
nonpedest ri
ancasual t
y '
'accidents,and
57%r eductionin``pedestri
ancasualty'
'accidents.

I
mpr
ovementi
nski
dresi
stance

Ifr
oadi sverysmoot ht
henskiddingofthev ehi
cl
esmayoccurori fthe
pavementi swetthenwetweat hercrashesoccurwhichaccountabout
20-30%.Thusi ti
simportanttoimprovetheskidresi
stanceoftheroad.
Variousway sofincr
easingtheskidresi
stanceofroadareby
construct
inghigh-f
ri
cti
onov er
layorcutti
ngofgroovesintothe
pavement .

Roadmar
kings

Roadmarkingsensur
epr operguidanceandcontr
oltothetraf
fi
cona
hi
ghway.Theyserveassupplement aryf
unct
ionofroadsign.Theyser
ve
aspsy
chological
barr
ieranddelineati
onoftr
affi
cpathanditslat
eral
clearancefrom t
raf
fi
chazardsf
orthesafemovementoft
raf
fi
c.Thus
theirpurposeistoprov
idesmoothandsafet
raf
fi
cfl
ow.

Gui
depost
swi
thorwi
thoutr
efl
ect
or

Theyarepr ov
idedattheedgeoft heroadwayt opr eventthevehicl
es
fr
om beingofft r
ackedfr
om t her
oadway .Theirprov i
si
oni sv
ery
essent
ialinhil
lyroadtopreventt
hev ehi
clefrom slidi
ngfrom top.Gui
de
postswithrefl
ectorgui
det hemov ementofvehicledur i
ngnight.

Guar
drai
l

Guar dr ailhavesi milarfunctionasofgui depost.Onhi ghembankment s,

hil
lyr oads, r
oadr unni ngpar all
el t
ot hebankofriver,shoresoflake,
near
rockpr ot r
usion,trees, bri
dge, abut mentsacoll
isionwi t
hwhi chisagreat
hazar df orav ehicle.Itisrequiredt oret
ainthevehicleont heroadway
whichhasacci dent all
yleftther oadbecauseoff aultorimproper
oper ati
onont hepar tofthedr iver.Dri
verwhohasl ostcont r
olcreat
ea
maj orpr oblem whi chcanbecur bedbyt hismeasur e.

Dr
iverr
evi
verst
op

Driverrev
iverstoparegenerall
yinusei ncountri
eslikeU.S.Awher e
driv
ercanst opandr efr
eshhimselfwi t
hf ood,
recreationandrest.They
playav er
yimpor tantparti
ntraff
icsafetyastheyrelievethedri
verfrom
thement altensi
onofconst antdri
ving.Thesestopsar erequi
redt obe
providedafterevery2hourtravelti
me.

Const
ruct
ingf
lyov
ersandby
pass

Inareaswher el
ocal t
raff
icishighbypassesar erequir
edt oseparate
thr
ought r
affi
cfrom localtr
affi
ct odecreasetheaccidentrate.To
minimizeconfli
ctsatmaj ori
ntersect
ionsflyover
sarer equir
edf orbet
ter
safetyandlessaccidentrate

Regul
aracci
dentst
udi
es

Basedonthepreviousrecor
dsofaccidentstheprev
enti
vemeasuresar
e
takenandaf
terthatthedatarel
atedtoaccident
sareagai
ncol
lectedt
o
checktheef
fi
ciencyofthemeasur esandforfut
ureimpl
ementat
ionof
f
urt
herpr
event
ivemeasur
es.

Saf
etymeasur
esr
elat
edt
oenf
orcement

Thevari
ousmeasuresofenfor
cementt
hatmaybeusefultoprev
ent
acci
dentsatspot
spronetoacci
dent
sareenumerat
edbelow.These
rul
esarerevi
sedfr
om ti
metotimetomakethem morecomprehensi
ve.

Speedcont
rol

Checksonspotspeedofallvehi
clesshoul
dbedoneatdiff
erent
l
ocationsandti
mi ngsandl
egalacti
onsonthosewhovi
olatethespeed
l
imitshouldbetaken

Tr
aini
ngandsuper
visi
on

Thetransportaut
hori
ti
esshouldbest r
ictwhil
eissuingli
cencetodr
iver
s
ofpubli
cservicevehi
clesandtaxis.Dr
ivi
ngli
cenceoft hedri
vermaybe
renewedafterspeci
fi
edperiod,onlyaf
terconducti
ngsomet estst
o
checkwhetherthedri
verisfit

Medi
cal
check

Thedri
ver
sshoul
dbet
est
edf
orv
isi
onandr
eact
iont
imeatpr
escr
ibed
i
nter
val
softi
me

Saf
etymeasur
esr
elat
edt
oeducat
ion

Thevari
ousmeasur
esofeducat
iont
hatmaybeusef
ult
opr
event
acci
dentsar
eenumerat
edbel
ow.

Educat
ionofr
oaduser
s

Thepassengersandpedestr
iansshouldbetaughttherulesoftheroad,
corr
ectmannerofcrossi
nget c.byi
ntr
oduci
ngnecessar yinst
ruct
ionin
theschool
sforthechi
ldr
enandbyt hehelpofpostersexhibit
ingthe
seri
ousresul
tsduetocarel
essnessofroadusers.

Saf
etydr
ive
Imposingtr
aff
icsafet
yweekwhentheroadusersarepr
operl
ydir
ected
bythehelpoftr
affi
cpoli
ceasameansoftrai
ningthepubl
i
c.Trai
ning
coursesandworkshopsshoul
dbeorgani
zedfordri
ver
sindif
fer
entpart
s
ofthecountr
y.

Saf
etyaudi
t

Itistheprocedureofassessmentoft hesaf et
ymeasur esempl oy edfor
theroad.Ithast headvant
agesl i
kepr operplanninganddeci sionfrom
beforehandensuresminimizati
onoffutureaccidents,t
helongt erm cost
associatedwithplanni
ngisalsoreducedandenabl esallki
ndsofuser s
toper cei
veclear
lyhow tousei tsaf
ely.Safetyaudittakesplacei nfive
stagesassuggest edbyWr i
sber
gandNi lsson,1996.Fi v
eSt agesof
SafetyAuditare:

1.Feasi
bil
i
tyStage-Thestart
ingpointforthedesigni sdet
ermi
ned
suchasnumberandt y
peofi nt
ersect
ion,r
elat
ionshipofthenew
schemetotheexi
sti
ngroad,therel
evantdesi
gnst andar
ds.

2.Draf
tSt age-Int hi
s stage hor
izontaland v
erti
calali
gnment
,
j
uncti
onl ay
outaredeter
mined.Aftert
hecompleti
onofthisst
age
deci
sionaboutl
andacquisi
ti
onistaken.

3.Detail
eddesignstage-Signi
ng,marki
ng,l
i
ght
ing,ot
herr
oadsi
de
equipmentandlandscapi
ngaredet
ermi
ned.

4.Pre-
openi
ngstage-Beforeopeni
nganewormodifi
edr oadshoul
d
bedri
ven,wal
kedorcycled.I
tshoul
dbedoneatdi
ff
erentcondi
ti
on
l
ikebadweather
,dar
kness.

5.Monitoringoft her oadinuse-Assessmenti sdoneatt hefinal

stageaf terther oadhasbeeni noperati
onf orfew monthst o
deter
mi ne whethert he uti
l
izat
ion i
s obt
ained as i
ntended and
whetheranyadj ustmentt othedesignarerequir
edint heli
ghtof
theactual behavi
orofroadusers.
Concl
usi
on

This chapt erprovi

des an i mportantsubjectofhi ghwaysaf etyand
accidentstudies.Everyt
hingat r
affi
cengineerdoes,from fi
eldst udies,
planninganddesi gn;tocont roloperati
onisrel
atedt ot heprovisionof
thesaf etysystem forvehiculartr
avel.Thi
schaptergiv esani nsightof
how theanal ysisoftraf
fi
cacci dentcanbedonef rom thev i
ewpoi ntto
reducei tbydesigni
ngpropersaf etymeasure.

Basi
cconcept
sofRoadSaf
etyAnal
ysi
s

Roadsaf etyisamaj orconcer

nfort r
ansportat
iongoverni
ngagenciesas
wellasor di
naryci t
izens.I
n ordert o gi
vesaf edriv
ing suggesti
ons,
carefulanalysi
sofr oadwaytraff
icdatai scri
ticalt
ofindoutv ari
ables
thatarecloselyrel
atedtofat
alaccident
s.

Someanal ysisont hi
st opichasbeendonebyFacul tyandst udentsof
depar tmentof comput ersci ence,Mi chigen Univ er
sity.They used
stat i
sti
cs anal ysis and dat a mining al gori
thms on t he FARS Fat al
Acci dentdat asetasanat tempttoaddr esst hisproblem.I nthestudythe
fatalr ateandot herat tr
ibutesincludi
ngcol li
si
onmanner ,anddr unk
driverwer ei nvest i
gated.Associationr uleswer ediscov eredbyApr iori
algor i
thm,cl assifi
cati
onmodelwasbui ltbyNaï veBay escl assi
fi
er,and
clusterswer efor medbysi mpl
eK- meanscl usteri
ngal gorit
hm.Cer tai
n
saf etydrivi
ngsuggest ionswer emadebasedonst at i
sti
cs,associati
on
rules,classi
f i
cationmodel ,andcl
ust er
sobt ained.

Dataminingusesmanydif
fer
enttechni
quesandal
gor
it
hmst
odi
scov
er
therel
ati
onshi
pinl
argeamountofdata.

Associat
ion rul
e mini
ng is a popul
armet hodology t
oi dent
if
ythe
si
gnif
icantrel
ati
onsbetweenthedatastor
edi nl
argedatabaseandal
so
pl
aysav eryimport
antr
uleinf
requenti
tem setmini
ng.

Acl
assicalassoci
ati
onr
ulemini
ngmet
hodistofi
ndfrequenti
tem set
s,
whi
chisthemet hodweuset
oanaly
zet
heroadwaytr
affi
cdata.
Cl
assi
fi
cat
ioni
ndat
ami
ningcanbeusedt
ocl
assi
fyr
ecor
dsofunknown
cl
assl
abel
s.

TheNaï veBay estechniquei soneofthev er

yf ir
stbasicpr obabi
l
ity
‐
basedmet hodsforclassif
icati
ont
hati
sbasedonBay es'hy
pothesi
swith
presumpti
onofi ndependencebetweeneachpairofv ari
abl
es.Thestudy
usedFARSdat aset( Fat
alit
yDataReport
ingSy st
em) .Thedat awas
coll
ectedandused.

1.
Dat
aPr
epar
ati
on

Thedat a pr
eparat
ion isthef i
rststep.TheDat a prepar
ation was
performedbef oreeachmodelconst ruction.Al
lrecordswi thmi ssi
ng
val
uei nt hechosenattr
ibuteswereremov ed.Al
lnumericalval
ueswer e
conv er
tedt oNominalvalueaccordi
ngt othedatadicti
onaryinat t
ached
usergui de.Fatalaccidents were calculat
ed and binned intot wo
categories:hi
ghandlow.

Sev
eral
var
iabl
esar
ecal
cul
atedf
rom ot
heri
ndependentv
ari
abl
es.

2.
Model
i
ng

Fi
rstsever
alstat
isti
calchar acteri
sti
csarecal culat
edf r
om thedat
aset
toshow t hebasicchar acteristi
csoft hef atalaccidents.Thenapply
associ
ationrul
emining,clustering,andNaïveBay esclassi
fi
cati
ontofi
nd
rel
ati
onshipamongt heat t
ri
but esandt hepatterns.

3.
Resul
tAnal
ysi
s

Ther esult
sanalysi
si ncl
udedassoci
ati
onrul
esamongthevari
ables,
cluster
ingofst at
esint heUSAont hei
rpopul
ati
onsandnumberoffatal
accidents,andclassif
icati
onoftheregi
onsasbeinghi
ghorlowriskof
fatalaccident.

Thet
ool
usedf
oranal
ysi
sisWeka.

Fi
rst
lythenumberoff
atalacci
dent
swerefoundf
ortheent
ir
eyearand
maximum werei
nJULYandmi ni
mum i
nFEBRUARY.
Thentherelat
ionbetweenfat
alacci
dentsandot
herv ari
ablescol
li
sion
ty
pe,speedl i
mit,li
ghtcondi
ti
on,weat
hercondi
ti
on,r oadwaysurface
condi
ti
on.Then t hi
s st
udywas carri
ed outi
n dif
ferentstates and
compared.

ThenFat
ali
tyr
atewasl
i
nkedt
oEmer
gencyMedi
cal
Ser
vices(
EMS)
.

The Table Iusi

ng Associ
ation Rule Mi
ning and cl
assi
fi
cat
ion wer
e
gener
ated.Tabl
eIIusedApriorialgor
it
hm showedf at
alacci
dentswit
h
hi
ghdrinki
nganddri
vi
ngwerehi gher.

Classif
icat
ionwasdoneofthediff
erentf
atalacci
dentdat
arecor
dssi
ng
NaïveBay esClassi
fi
eranditshowedrel
evancetodependenceoff
atal
accidentsontheatt
ri
butest
aken.

Cluster
ingofst ateswasdonedependi
ngonthepopulat
ionandt he
accidentdatacoll
ectedandt
her
egi
onsweredi
str
ibut
edi
ntoblue,
green
andr edzone.

Empiri
calBay
esMet
hodofEst
imat
ingSaf
etybyt
heEmpi
ri
calBay
es
Method

1.
Int
ruduct
ion

Empri
calBay
esmethodi
susedf
ort
heCompr
ehensi
veHi
ghwaySaf
ety
I
mprovementModel
(CHI
SM)

ACasest
udyhasbeent
akent
ost
udyt
hemodel
.

Thesaf et
yofanent i
ty(aroadsection,aninter
section,adriver,abus
fl
eet etc.
)i s defined by the number of acci dents,or acci dent
consequences which are expect
ed to occuron t he enti
tydur ing a
speci
fiedperi
od.Since,whati sexpectedcannotbeknown,saf et
ycan
onlybe estimated,and est i
mati
on isi n degr
ees ofpr ecision.The
preci
sionofanestimateisusuall
yexpressedbyitsstandarddev i
ati
on.

Thesafet
yofent i
ti
esonwhi chmanyaccidentsoccurduringashor
t
peri
odcanbeestimatedpr
eciselybyusi
ngonlyaccidentcount
s.I
fon
theroadoneexpects100accident
speryear
,then,withthr
eeyear
sof
accidentcounts,
onecanest i
matetheav
erageyear
lyacci
dentf
requency
withstandarddeviat
ionofabout(
100̸3)=±5.7accident
s̸ yearor5.7
%oft hemean.( Thi
si sbasedontheassumpt
ionthatacci
dentcounts
arepoissondistr
ibut
ed).

Conv ersely
,wheni tt akesal ongt i
mef orf ew acci
dent st ooccur ,the
estimatei si mprecise.Thus,e. g.,ifoneexpect sar ai
l-highwaygr ade
crossingoradr i
vert ohav eoneacci denti nt enyearst hen,wi t
ht hree
years ofacci dentcount s,the est i
mat e ofav er
age y early accident
fr
equencyhasast andar ddev iat
ionof( 0.1̸3) =±0. 18.Meani s0. 1
accidentspery eart hest andar ddev iati
oni s180% oft hemean.Thus,
oneshor t
comi ngofsaf etyest i
mat est hatar ebasedonacci dentcount s
onlyist hattheymaybet ooi mpreciset obeusef ul
.Short comingoft hi
s
analysisist hatitissubj ectedtocommonbi as.Attent
ionneedst obe
givent otheent iti
eswhi chmayl eadt ohi ghernumberofacci dents,or
mer i
tat t
entiontot hem becauseoff eweracci dent
s.Regressiont omean
analysishasn’ tgivenaccur ateresultst hust hechangei nmet hodof
analysisisrequired.

TheEmpi r
icalBay
es( EB)met hodfortheestimati
onofsafetyi
ncr
eases
theprecisi
onofest i
mationandcorrect
sf ortheregr
essi
on-to-
meanbias.
Thet heoreti
calfr
amewor kf orcombiningtheinformat
ioncontai
nedin
acci
dentcount swi t
htheinformati
oncont ai
nedinknowingthesafet
yof
simi
larentiti
esistheEBmet hod.

Appl
i
cat
ionsofEBmet
hodi
ncl
ude

 Usagei
nInt
eract
iveHi
ghwaySaf
etyDesi
gnModel
(IHSDM)

 Usagein Compr
ehensi
veHighwaySaf
etyI
mpr
ovementModel
(CHSI
M)nowunderdevel
opment.

I
tshouldbeusedwher evertheneedforroadsafetyari
ses,whetherin
t
hesearchforsit
eswithpromise,t
heevaluati
onofthesafet
yeffectsof
i
nter
venti
ons,ortheassessmentofpotenti
alsaf
etysavi
ngsduet osite
i
mprovements.

2.
TheEBPr
ocedur
e

Thetaski
stomakejoi
ntuseoftwocl
uestothesafet
yofanentit
y:t
he
acci
dentr
ecor
dofthatenti
tyandtheacci
dentf
requencyexpect
edat
si
milarentit
ies.Thi
sexpectedaccidentrecordofthatent
ityandt he
acci
dentfrequencyexpect
edatsimilarent
it
ies.Thi
sexpect
edaccident
fr
equencyatsi mil
arenti
ti
esisdeterminedbyt heSafet
yPerfor
mance
Functi
on(SPF) .

Est
imat
eoft
heExpect
edAcci
dent
sforanent
it
y=

Weight×Acci
dentsexpect
edonsi
mil
arent
it
ies+(
1-Wei
ght
)×Count
ofaccident
sont
hisent
ity

Ther esulti
sdet erminedbyhowmuch‘ weight’i
sgivent otheaccidents
expectedonsi mi l
arenti
ti
es.Thestrengt
hoft heEBmet hodisintheuse
ofa‘ weight’t
hatisbasedonsoundl ogi
candonr ealdata.This‘
wei ght
’
wil
lbeseent odependont hestrengthoftheaccidentrecord(howmany
accidentsar etobeexpect ed),andont her eli
abil
i
tyoft heSPF( how
dif
ferentmaybet hesafetyofaspecifi
csitefrom t
heav eragewhicht he
SPFr epresents)
.

TheEB est i
mat i
onpr ocedurecanbeabr idged orfull.Theabr i
dged
versionmakesuseoft herecent2-3y earsofaccidentcount sandoft he
averaget r
af f
icflowf orthatperi
od.Thi srefl
ectsthenowcommonbel i
ef
thataccidentcount st hatareol derthan2- 3y earsmaynotr epresent
currentcondi ti
ons.Howev er,theEBpr ocedurer emovesmostr easons
fornotusingol derdat a.Accordingly
,thef ul
lversi
onoft heEBpr ocedure
makesuseofal ongeraccidentandt raffi
cflowhistor
y.Becauset hefull
procedureusesmor eaccidentcount s,theestimateofthef ul
lprocedure
i
smor epr eciset hantheest i
mat eproducedbyt heabridgedprocedur e.
Ther ef
ore,ifdatai savail
able,
oneshoul dst r
ivetousethef ullpr
ocedure.

3.SAFETYPERFORMANCEFUNCTI
ONANDWEI
GHT.

The av erage accidentf requencyof‘ simil

arsit
es’and t he vari
ati
on
aroundt hisav eragear ebr oughti ntot heEBpr ocedurebyt heSaf ety
PerformanceFunct i
on( SPF) .TheSPFi sanequat i
ongivi
nganest i
mat e
of µ, t he av erage acci dents/ (
km- y ear
)f or road segment s or
accidents/yearforintersecti
ons, asaf unctionofsomet r
aitvalues(e.g.
,
ADT, Lanewi dth,...)andofsev er alregressionparameter
s.
 Toill
ustrate,consi
dertheSPF:est i
mat eofµ=0. 0224×ADT0.564fora
cert
ainkindofr oadinagivenjuri
sdict
ion.HereADTpl aystherol
eofone
tr
ait
sv alue,noaddi ti
onaltrai
tvaluesar erepresentedi ntheSPF,the
esti
mateofoner egressi
onparamet eris0.0224,andt heesti
mateofthe
secondregr essi
onpar ameter0.
564.Ifonar oadoft hiskindADT=4000
vehi
cles perday ,then one shoul d expect0. 0224×40000.564=2.
41
acci
dents/( km-year)
.

SPFsar ecalibratedf rom dat abyst at i

sticaltechni ques.I n
thepasti twascommont oassumet hatacci dentcount scome
from aPoi ssondi str
ibut i
on.Howev er,resear cher sfoundt hatt he
acci dentcount susedi nt hecal ibrati
onofSPFsar eusual lymor e
wi delydi spersedt hanwhatwoul dbeconsi stentwi tht hePoi sson
assumpt ion.Thi siswhyi tisnowaday scommont oassumet hat
theacci dentcount swhi chser veasdat acomef r
om anegat ive
binomi aldistri
bution.Oneoft hepar amet er soft hisdist ri
butioni s
the‘ ov erdispersionpar amet er’
,denot edher eby‘ ν’
.Forr oad
segment s,theov erdi sper sionpar amet erisest i
mat edper -unit-
l
engt h.Thati s,thedi mensi onofνi s[1/ km]or[ 1/mile].The
meani ngofνcomesf rom t hef oll
owi ngr elationship:ifLi st he
l
engt hofasegmentand0i st heexpect ednumberofacci dent sfor
thatsegment ,
thent hev arianceofacci dentcount sonsegment s
oft hatki ndi s0[1+0/ (νL)].Thedi mensi onsofνandLmustbe
compl ement ary.Thati s,ifint hecour seofmodel calibrationνi s
est imat edperkm, t
henLmustbemeasur edi nkilomet r es.Not e,ν
est imat edperkm =0. 622×νest i
mat edpermi le.Forint ersect i
onsL
i
st akent obeone.Mor edet ailandanexpl anat i
onoft hesour ces
ofov erdi spersi
oni sinr ef erence.

µ Thenumberofacci dent
s/(km-y
ear
)forexpectedonsi
mi l
ar
segment
sandacci
dent
s/yearexpect
edf
orsimil
arinter
sect
ions.

O thenumberofaccident
sdur i
ngaspecif
iedperi
odgivenby
µ×L×Yexpectedforsi
mil
arsegmentsandµ×Yexpectedforsi
milar
i
ntersect
ions.I
nthi
sLstandsforsegmentl
engthandYforyear
s.

ν overdispersi
on parameteresti
mated perunitlengthfor
segments.Nat
urall
y,ent
it
iesforwhichtheaccidentf
requencyis
notproport
ionaltothei
rlength(
e.g.i
nter
sect
ionsorrail
-hi
ghway
gradecrossings)haveanov erdi
spersi
onparameterthatisnot
esti
matedperunitlengt
h.

I
tisnowpossi
bletogi
vet
heexpr
essi
onf
ort
he‘
wei
ght
’used
i
nequat
ion1.I
ngeneral
:

.
..2

4.THEABRI
DGEDEBPROCEDUREI
LLUSTRATED.

To i
ntr
oduce the abri
dged pr
ocedure consi
dernumer
ical
exampl
esofgradual
l
yincreasi
ngcomplexi
ty:

NumericalExampl
e1:ARoadsegmentwi
thoney
earofacci
dent
count
s.

Ar oadsegmenti s1.8km long,hasanADTof4000,and

recorded12accidentsi
nthelasty
ear .TheSPFforsimi l
arroadsis
0.0224×ADT0.564 acci
dents/
(km-year)
,with an overdi sper
sion
parameterν=2.05/km.Toesti
matet hesaf
etyofthi
sr oadsegment
proceedasfoll
ows.

St
ep1:
Aver
agef
orent
it
iesoft
hiski
nd.

Roadssuchasthishave0.0224×40000.
564=2.
41acci
dent
s/
(km-year
),onaver
age.Therefor
esegment sthatare1.8km l
ong
areexpectedt
ohave1.8×2.
41=4.34accident
sinoneyear
.
St
ep2:
Wei
ght
.

Weneeda‘ wei
ght’forjoi
ningthe12accidentsrecordedon
thi
sroadandt he4.
34acci dentsforanaver
ager oadoft hi
skind.
Forweightweuseequat ion2.Her eµ=2.
41acci dent
s/(km-year
),
Y=1 and t heestimateofν =2.05/
km.Therefore:weight=1/
[
1+(2.41×1)
/2.05]=0.460.Not ethatbothµ andνar e‘ perunit
l
ength’.

St
ep3:
Est
imat
e.

Usingequat ion1t heestimateoft heexpect

edaccidentf r
equency
for t he speci f
ic r oad segment at hand i s:
0.460×4. 34+0.540×12=8.48acci dent
sinoney ear
.Notet hat8.48is
betweent heav erageforsimil
arsites(4.
34)andtheacci dentcount
fort his site( 12).The EB est imatorpull
st he accidentcount
towar dst hemeanandt herebyaccount sforther egr essi
ont o
meanbi as.Thest andar
ddev iat
ionoftheesti
mateoft heexpect ed
accidentf requencyi sgi
venby :

…….3

Her
e,ϕ=±̸ (
0.54×8.
48)
=±2.
14acci
dent
sinoney
ear
.

Numer
ical
Exampl
e2:
Thr
eey
ear
sofacci
dentcount
s

Supposenowt hatfort
hesamer oadsegmentwehav ethree
yearsofacci dentcount
s:12,7,8,andt hattheADTi neachof
thosethr
eey earswas4000vpd.Toestimatethesaf
etyoft
her oad
segment:

St
ep1:
Aver
agef
orent
it
iesoft
hiski
nd.

Asbefor
e,segmentsoft
hiski
ndareexpectedtohav
e2.41
accident
s/km-year
. On 1.8 km in thr
ee y ears we expect
1.8×3×2.
41=13.01acci
dent
s.
St
ep2:
Wei
ght
.

Thewei ghti
s1/ [
1+(2.
41×3)/
2.05]
=0.220.Notethatwi
thoney ear
ofaccidentdatausedt heweightwas0. 460.Asmor ey ear
sof
acci
dentdat a as used,the weight(given t
ot he numberof
acci
dentsexpectedonsimil
arenti
ti
es)dimini
shes.

St
ep3:
Est
imat
e.

Expectedacci
dent
s=0.220×13.
01+0. 780×(
12+7+8)=23.92acci
dent
sin
thr
eey ear
swithϕ=±̸( 0.
78×23.92)
=±4.32or
23.92/
(3×1.8)
±4.
32/(
3×1.8)=4.
43±0.80acci
dent
s/(km-year)
.

Numer
ical
Exampl
e3:
Appl
i
cat
ionofAcci
dentModi
fi
cat
ionFunct
ions
(
AMFs)

Supposenowt hattheSPFequat
ioninExample1isforroadswit
h
1.5m shoulderswhil
etheroadsegmentofint
eresthas1.
2m shoulder
s,
andthata0. 3m decr
easeinshoul
derwidt
hisknownt oi
ncrease
acci
dentsby ,say,
4%.

St
ep1:
Aver
agef
orent
it
iesoft
hiski
nd.

Usingther esul
tfrom Example1,segmentsofthi
ski
ndareexpect
edto
have1.04×2.41=2.51accident
s/km-year
.On1.8km i
nthr
eeyearswe
expect1.8×3×2.51=13.
55accidents.

St
ep2:
Wei
ght
.

Thewei
ghti
s1/
[1+(
2.51×3/
2.05]
=0.
214.

St
ep3:
Est
imat
e.
Expectedaccidents=0.214×13.
55+0.
786×(12+7+8)=24.12±
̸(0.786×24.12)=4.35accidentsi
nthr
eey ear
sor
[
24.12±4.35]
/(3×1.8)=4.47±0.
81acci
dents/(
km-year).

Numer
ical
Exampl
e4:
Subsect
ionsandAcci
dentr
ecor
ds.

Considerthe r
oad segmentin Fi
gure1t hatis made up oft hree
subsecti
onsthatdif
ferinsomet r
ait
s(whichdet erminet hevari
able
val
uesoft heSPF)andintheAMFs.However,t
heacci dentcountisnot
avai
lablesepar
atel
yforeachsubsecti
on,onlyf ortheent ir
e1.5km
segmentonwhi ch11acci
dentswer
ecountedinthelasttwoy ear
s.

Fi
gur
e1

St
ep1:
Aver
agef
orEnt
it
iesoft
hiski
nd.

TheADTsandAMFsdi
ff
eramongstt
hesubsect
ionsasshowni
n
col
umns2and4ofTabl
e1.
Tabl
e1

Subsect
io ADT Lengt
h AMF Acci
dents/(
km- Acci
dent
s
n [km] year)
1 2000 0.1 .
90 1.466 0.
147
2 2300 1.2 .
95 1.675 2.
010
3 2300 0.2 1.05 1.851 0.
370
Sum 2.
527

Assumet hat,asintheearli
erexamplestheSPFi s0. 0224×ADT0.564
accident
s/(km-year)andν=2.05/km.Thus, af
tercorrectionfor
0.564
AMF, subsecti
on1i sexpectedtohav e0.0224×2000 ×0. 90=
1.466accidents/(
km-year)andtherefor
e1.466×0.1=0.147
accident
s/year.Thethreesub-sect
ionstogetherareexpect edto
have2.527×2=5.054accidentsintwoy earsor2.527/ 1.5=1.715
acci
dents/(
km-year
).From her
eoniti
sconvenientt
ofor
getabout
thesubsect
ionsandtreatt
he1.5km segmentasoneenti
ty.

St
ep2:
Wei
ght
.

Thewei
ghti
s1/[
1+(
1.715×2)
/2.
05]
=0.
374.

St
ep3:
Est
imat
e.

Expectedaccidentsforthe1.
5km l
ongsectionintwoyears
=0.374×5.054+0.626×11=8.78±̸ (
0.626×8.78)=2.
34acci
dent
s
or[8.78±2.34)
/(1.5×2)=2.
93±0.
78accident
s/(km-year
).