You are on page 1of 16

TrafikdagepAalborgUniversitet2012 ISSN16039696 1

Reliability of Railway Operation


Alex Landex, al@transport.dtu.dk
DTU Transport

Abstract
Delaysinrailwaynetworkscauseproblemsinthedailyoperationandresultinreducedreliabilityofthe
railwayoperation.Thearticledefinesmethodsbasedonthetwocloselyrelatedconceptsofregularityand
punctuality.Thesemethodscoverthetraditionalmeasuresfortrainsbasedontherealizedandsimulated
operationbutalsomoreadvancedmethodsthattakethepassengersexperienceintoaccount.Thisisdone
byusingpassengerdelaymodelsthatestimatesthedelaysofthepassengersbasedontraindelaysand
knowledgeaboutthepassengerbehavior.Forhighfrequentoperation,thetraditionalmeasurescannot
describetheperceivedservicereliabilityofthepassengersaspassengerscantravelwithdelayedtrains
withoutknowingofthedelay.Here,thearticlepresentsotherreliabilitymeasures.Usingthepresented
reliabilityapproaches,thearticleillustrateshowthereliabilityofrailwayoperationcanbeimproved
throughreliabilityanalyses.

1 Introduction
Delaysinarailwaynetworkareoneofthebiggestproblemsinthedailyoperationsofarailwaycompany
(Berger,Hoffmann,Lorenz&Stiller2011).Therefore,attentiontotransitqualityandefficiencyingeneral
andreliabilityinparticularisincreasing.(vanOort,vanNes2010).Forrailwayservices,reliabilitycanbe
definedasthecontinuityofcorrectservice(Avizienis,Laprie,Randell2001).Seenfromthepassengers
pointofview,thereliabilitycanbedividedintothetwocloselyrelatedconceptsregularityandpunctuality.
Regularityisthevariationinheadwayswhilepunctualityrelatestothedeviationfromthescheduledarrival
anddeparturetimes(vanOort2005).

Trainsarenotallowedtodepartbeforetimesincefromapassengerperspective,thenextdepartureis
latebythefrequency(Bush2007).Therefore,thetrainswillalwaysdepartontime/punctualor
delayed.Ifthetimeisdividedinsmalltimeintervals(e.g.seconds),itisverydifficultforthetrainstodepart
exactlyontime.Sincethetrainsarenotallowedtodepartbeforetime,theriskof(very)smalldelaysishigh
(Landex2009).Toavoiddiscussionsofwhenatraindeviatesfromitstimetable(i.e.,thetrainispunctual),
mostcountrieshavedecidedonthresholdvaluesforwhenatrainisontime.InDenmarkitisdecidedthat
thetrainsarepunctualiftheyarrivewithinthefollowingthresholds(Landexetal.2007;Landex2008):
Strains2minutes
Regionaltrains5minutes
IntercityandIntercityExpresstrains5minutes
Freighttrains10minutes
Denneartikelerpubliceretidetelektronisketidsskrift
ArtiklerfraTrafikdagepAalborgUniversitet
(ProceedingsfromtheAnnualTransportConference
atAalborgUniversity)
ISSN16039696
www.trafikdage.dk/artikelarkiv
TrafikdagepAalborgUniversitet2012 ISSN16039696 2

Thethresholdvalueforwhenatrainispunctualdiffersbycountry.IntheUSA,thethresholdvaluevaries
from5to30minutesdependingonthelengthofthetrainserviceandnottheaveragetraveldistancefor
thepassenger(Bush2007).InEurope,thethresholdvaluesvarytoo,butalsowheretomeasurethe
punctualityvaries.E.g.TheNetherlandsmeasurethepunctualityatthedeparturefrom32measuring
stationsspreadoutoverthenetwork,whileGermanyandNorwaymeasurethepunctualityatarrivalatthe
terminalstation
1
(Daamenetal.2007;OlssonandHaugland2004).Becauseofthedifferentwaysof
measuringpunctualityonlyfewinternationalcomparrisonsoftraindelays(e.g.(NederlanseSpoorwegen
2001;Vromans2005))areconducted,cf.Figure1.

Figure1:Trainsarrivinglessthan5minuteslateinyear19992002.Datafrom(NEA2003).

Inhighfrequencysystems,morefocusonheadwaydeviationiscommonbecausetravelerstendtoarriveat
random(vanOort,vanNes2010).However,incasethevehiclesoperateslowerthanplanned,theheadway
deviationmethodmaynotbesufficienttodescribethereliabilityoftherailwayoperation.Toachievea
methodwithlowcomplexity,theheadwaydeviationcanbecombinedwiththerunningtimedeviationof
thevehicles(Landex2009).

Toguesstimatetheimpactonthereliabilityoftherailwayoperationincaseofalteredtimetablesor
modifiedinfrastructure,itisnecessarytoconductasimulationoftherailwayoperation.Therefore,this
articlefirst(insection2)presentsthetraditionalmethodofsimulatingtraindelays.Thisis(insection3)
followedbyanapproachthattakespassengerdelaysintoaccountwhenmeasuringthereliabilityofthe
railwayoperation.Insection4,reliabilitymeasuresforhighfrequenttrainoperationarepresented.Before
theconclusions(section6),section5discuseshowthereliabilityofrailwayoperationcanbeimproved
basedonthedescribedmethods.

2 Simulation of Train Delays


Thedelaypropagationandthetotaldelaycanbecalculatedforagiveninitialdelayprovidedthatthe
timetableandthecharacteristicsoftheinfrastructureareknown.However,thesecalculationsofdelaysare
onlypossibleforidealizedsituations.Thisisbecauseonlyonerailwaylinewouldbeexamined,no
dispatchingwouldbeincluded,anditwillbeassumedthatthetimetablewillregeneratebeforethenext
delayoccurs.However,adelayononerailwaylinemayinfluenceotherrailwaylinestoo,andthetotal
amountofdelaymaybereducedthroughdispatching(e.g.changingtheorderofthetrains).Furthermore,

1
In Norway, punctuality is measured at some importat stations too.
70% 75% 80% 85% 90% 95% 100%
Finland
Hungaria
Switzerland
Denmark
Austria
Belgium
Netherlands
Norway
Sweden
Long distance trains
Local trains
All trains
TrafikdagepAalborgUniversitet2012 ISSN16039696 3
iftwoinitialdelaysoccurshortlyaftereachother,theseconddelaymightnothavethesameinfluence,as
thetrainwouldhavegotaconsecutivedelay,alsoifthedelaydidnotoccur,cf.Figure2.

Figure2:Twoinitialdelaysshortlyaftereachother(brokenlinesaretheplannedtimetable).

Insteadofcalculationsonidealizedsituations,simulationcanbeused.Originally,thescienceofsimulation
wasthereproductionofarealobjectoraprocessasamodel.Inasimulation,thismodelwasusedinstead
oftheoriginal.However,giventhesuccessofcomputersasatechnicaltool,almostallprocessescannow
besimulatedwithcomputerprograms.Duringrecentyears,ithasbecomeclearthatsimulationisasuitable
methodtoreproducetherealityinavirtualprocess.Theresultscanhelptounderstandandanalyze
processesmoreeasily(Siefer2008).

Simulationmodelsareoftenmoreprecisethantheidealizedformulas.Thisisbecausesimulationcantake
delaysoccurringimmediatelyaftereachotherintoaccountandbecausesimulationmodelsoftenhave
detailedknowledgeabouttheinfrastructureandtrainoperationsothattheycan,forexample,calculate
thedelaycausedbyspeedreductionsduetorestrictivesignals.Furthermore,morerailwaylinescanbe
examinedatthesametime,dispatchingcantosomeextentbeincluded,andpreviousdelaysareincluded
whenthetotalamountofdelayiscomputed.Therearemanysimulationmodelsthatcanbeusedfor
differentanalysisbutingeneral,themodelscanbedividedintothreecategories,cf.Figure3.

Foranalysingdelaysanddelaypropagation,tactical(and/oroperational)modelscanbeused(e.g.RailSys
(Siefer,Radtke2005)andOpenTrack(Nash,Huerlimann2004))
2
.Thesetypesofsimulationmodelsare
generallybuiltupinseveralsteps.First,theinfrastructuremustbebuiltupbeforeconstructingthe
timetable.Then,thedelaydistributionoftheinitialdelaysmustbeenteredtogetherwiththerulesfor
dispatching.Finally,thesimulationcanberunandtheresultscanbeevaluated,cf.Figure4.Toensurea
stableandreproducibleresult,50200simulationsshouldbeconducted(Siefer2008)
3
.

2
(Barber et. al 2007) and (Koutsopoulos, Wang 2007) give an overview of different simulation tools.
3
(Radtke, Bendfeldt 2001) suggest 50100 simulations.
Time
Distance
Point of delay
Point of delay
TrafikdagepAalborgUniversitet2012 ISSN16039696 4

Figure3:Differenttypesofsimulationmodelsandtheirmaintypesofanalysis.Basedon(Kaas1998).

Theadvantagesofsimulationmodelscomparedwiththeanalyticalmodelsarethehighaccuracyandthe
possibilitytotestfuturechangesintheinfrastructureandthetimetablesforanentirenetwork.
Furthermore,simulationmodelscanestimatetheimpactofthetrainsinthecaseofreducedspeedand
queuing,skipanyconditionalstops,andchangingtheorderofthetrainstoreducetheamountof
consecutivedelays.However,simulationmodelsalsohavetheirlimits.Althoughsimulationmodelsare
moreaccuratethanidealizedformulas,theycannotdispatchthetrainsinthesamewayasdoesthe
dispatcherinreallifewheremoretoolstorelievethedisruptionsexist,e.g.,(Jacobs2008,JespersenGroth
etal.2007):
Overtaking
4
:Choosetoovertakeanothertraintohavebetteroverallreliefinthedisruptions
Changeinstoppattern:Adelayedtraincanskipsomestationstorecoverfromitsdelay.Another
trainmightthenhaveadditionalstops.ThiskindofdispatchingisoftenusedontheCopenhagen
suburbanrailwaynetwork
Insertinganontimetrain:Inthecaseofdelays,thedelayedtraincanbereplacedbyanothertrain
runningontime.ThiskindofdispatchingisoftenusedontheCoastline(fromCopenhagento
Elsinore)inDenmark
Increasingresidualcapacity/canceltrain(s):Cancellationofoneormoretrainscanensure
sufficientcapacitytoavoidtoomanyconsecutivedelays.Inthisway,thenumberofdisruptionscan
bereduced.Thecancellationoftrainscanbeovertheentirerouteoronlypartofit.Thiskindof
dispatchingisoftenusedontheCopenhagensuburbanrailwaynetwork
Useofalternativeroutes:Ifthereareparallelroutesbetweentwostationsalongarailway
corridor,thetraincanchooseanalternativeroute.Itmightbenecessarytoskipsomestationsto
beabletochangetoanothertrainservice.ThiskindofdispatchinghasbeenusedinDenmarkfor
freighttrainsinthecaseofmajorincidents
Bundletrains:Inthecaseofreducedcapacity(e.g.duetoaclosedtrack),extracapacitycanbe
gainedbybundlingthetrains.Thetrainsinthesamedirectioncanbebundledbyspeedorby
directionforunscheduledsingletrackoperation.ThiskindofdispatchinghasbeenusedinDenmark
incasesofconstruction/maintenanceworkandincasesofunscheduledclosureoftracks

4
Simulation tools often allow unscheduled overtakings in the case of disruptions but the dispatcher might choose the overtaking based on criteria
other than the dispatch algorithmand possibly even combine the overtaking with other dispatching actions.
Operational models
Tactical models
Strategic models
Planning horizon
L
e
v
e
l

o
f

d
e
t
a
i
l
High
Low
Short Long
Operation
Special
timetables
Timetable
Plan of operation
Traffic forecasts
TrafikdagepAalborgUniversitet2012 ISSN16039696 5
Couplingtrains:Insomecases,twoormoretrainscanbecoupledtoonetrain.Inthiswayfewer
trainshavetopassabottleneckandcapacityisgained.Thiskindofdispatchinghasbeenusedin
Denmarkincasesofconstruction/maintenanceworkandincasesofunscheduledclosureoftracks

Figure4:Typicalstepswhensimulatingdelaysinrailwaynetworks(Landex,Nielsen2006a).

Besidesthedispatchingrules,thedifficultyincalibratingsimulationmodelscomesfromthedifficultlyin
enteringtherightinitialdelays,becauseitisdifficult,orevenimpossible,tofindastandarddistribution
typethatiswidelyapplicable(Yuan2006).Tocalibratethemodelstheinitialdelayscanbeextracted
5
from
collecteddataabouttheactualoperation
6
(Tromp2004).Byidentifyingtheconsecutivedelays
7
,itis
possibletoderivetheinitialdelayfromtheactualoperation.Theseinitialdelayscanbeenteredinthe
simulationmodelandthefinalcalibrationcanbeconductedbasedontheoutputs,cf.Figure5andre
runningthestepsinFigure4.Thiscalibrationcanbetimeconsuming(Kaas2000)asevensmallchangesin
thedelaydistribution(s)canresultinchangeselsewhereinthenetwork.

Whenusing(microscopic)simulationmodelstopredicteffectsofscenariositisnecessarytocalibratethe
simulationmodelstogivetheresultsactuallyobserved.However,theliteratureoncalibrationofrail
simulationmodelsislimited(Koutsopoulos,Wang2007).Thedifficultycalibratingsimulationmodelstogive
exactlythesameresultsasinreallifeoperation,resultsinadhocadjustmentsbasedonsimplestatisticsor
performancemeasurestocomparethesimulationoutputtofieldobservations.Here,theadjustingofthe
modelparametersisoftendonebytrialanderror,orsomekindofestimation,untilthesimulated
measuresareclosetotheobservedones,cf.Figure5.Oftenthepurposeofcalibratingsimulationmodelsis

5
It should be noted that the observed delays are the sumof the initial delays and the consecutive delays.
6
(de Fabris, Longo & Medeossi 2008) describe a method to analyze the actual operation.
7
(Daamen, Goverde & Hansen 2007) describe a method to identify consecutive delays.
95,4%
84,0%
15,7%
90,6%
80,5%
17,3%
95,4%
90,3%
22,5%
91,4%
86,8%
22,5%
92,7%
85,0%
19,6%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Morning Day Afternoon Other time Total
Train regularity [%] Passenger regularity [%] Arrivals before time [%]
Infrastructure
Timetable
Evaluation
Simulation
TrafikdagepAalborgUniversitet2012 ISSN16039696 6
toreproducetheoperationofanaverageday,tobeabletoexaminetheconsequencesofchangesinthe
operationand/orinfrastructure.

Figure5:Calibrationofsimulationmodels.Basedon(Toledo,Koutsopoulos2004).

Arobusttimetableisabletodealwithonlyminorinitialdelaysafewminutesofdelaybecauseno
reasonabletimetableissufficientlyrobustinthecaseoflargedelays(Vromans,Dekker&Kroon2006).
Perturbedoperationcausesdelaysand,atleastincongestedareas,thisrequiresfrequentrealtime
modifications(D'Ariano2008).Thisisfurthercomplicatedbythelimiteddispatchalgorithms,whichmakes
thecalibrationevenmorecomplicated.

Microscopicsimulationmodelsperformaccordingtotheappliedalgorithmsandtherulesentered.
However,inthecaseofdelaysmorepassengersmightwanttousethesametrain,whichmightresultin
longerdwelltimes.Thismeansthatdelayedtrainsgenerallyhaveahigherriskofbecomingevenmore
delayed.Thisisgenerallynottakenintoaccountinmicrosimulationmodels.However,ifanothertrain
goinginthesamedirectionhasovertakenthedelayedtrainandisrunningimmediatelyinfrontofthe
delayedtrain,thedelayedtrainmighthavefewerpassengersandtherebypossibleshorterdwelltimes,
whichiswhythetrainmaypickuptimefaster.

Althoughsimulationmodelshavedisadvantages,theyaremoreaccuratethansimpledelaycalculations.To
overcome,oratleastreduce,thedisadvantagesofsimulationmodels,theresultsareoftencompared
relativelytoeachother.Inthisway,itispossibletoexaminetherelativedifferencesfordifferentprojects
andchoosethebestalternative.

Microscopicsimulationmodelsarepowerful,buttheyrequireextensiveworktoenterthedetailed
infrastructuretopology,traincharacteristics,signallocationsandtimetables.Furthermore,thesemodels
requiremorecomputingtimecomparedwithsimplermodels,whichiswhyitmightbetemptingtouse
thesesimplermacroscopicoroperationalresearchmodels
8
.Althoughthesemodelsaresuitablefor
evaluatingtheoverallstabilityoftimetablesofinterconnectedlines,theycannotbeusedtoestimatethe
distributionsofconsecutivedelaysandthepunctualitylevelofthescheduledtrains,astheyaregenerally
basedonadeterministicmodelingapproach(Yuan2006).Nevertheless,thesimplermodelsgivean
indicationaboutthedelay(s)thatmightbesatisfactoryatleastinascreeningprocesswithseveral
alternatives.Toachievebettersimulationresultsfromsimulationmodels,furtherresearchand
developmentisneeded,especiallywithindispatchingalgorithms.

8
Microscopic models require more detailed data than do the simpler models. In some countries these data are difficult to procure, but in Denmark
there is open access to the necessary data.
Simulation
Output
Simulation
Output
Real
Output
Real
Output
Simulated
System
Simulation
Simulated
System
Simulated
System
Simulation
Real
System
Operation
Real
System
Real
System
Operation
C
o
m
p
a
r
i
s
o
n
Real
Input
Initial delays
Real
Input
Real
Input
Initial delays
Simulation
Input
Delay distribution
Simulation
Input
Simulation
Input
Delay distribution
E
s
t
im
a
t
io
n
Model
Real life
TrafikdagepAalborgUniversitet2012 ISSN16039696 7

3 Passenger Delays
Simulationoftraindelaysdoesnotreflectthepassengersperceptionofreliability.Thiscaninsteadbe
donebycalculatingtraveltimeforpassengersfortheplannedsituationandtheactuallyperformed
operation.Thesetimescanthenbecompared,andtheextendedtraveltimeofthepassengersor
passengerdelaycanthenbeusedasameasureforthereliabilityoftheoperation.

Severalmethodshavebeenappliedtocalculaterailpassengerdelays.Themajorityofthesedonotinclude
schedulebasedroutechoicemodels;inthefollowingclassifiedas0
th
generationmodels.Somemethods
aresimpleandstraightforwardtoimplement,butalsosimplifiedandinaccurate,whileothermethodsare
moreprecise.Thesimplestmethods(0
th
generation)arenotreportedintheinternationalliterature,asthey
havebeendevelopedandusedbyrailwaycompanies(Landex2008).Althoughthe0
th
generationmethods
aresimpleandinaccurate,thesemodelsfoundthebasisforthelater1
st
,1,2
nd
and3
rd
generation
passengerdelaymodels(basedon(Landex2008)):
0
th
generationmodelsarethesimplestmodelswherethepassengerdelaysareassumedequalto
thetraindelaysorthetraindelaysmultipliedwitheitheramountofpassengersinthetrain(cross
sectiondelays)ortheamountofalightingpassengersatthestations(countingtraindelays).These
modelsdonottaketransfersandchangedroutes/trainsduetodelaysintoaccount.
1
st
generationmodelscomparethetimethepassengersuseinthesystemcalculatedbyanoptimal
routechoicemodelbasedontheplannedtimetableandtheactuallyperformedoperation.These
modelstakethepassengersentirerouteintoaccountbutthemodelsassumethatthepassengers
knowalldelaysinadvance.
1generationmodelsaresimilarto1
st
generationmodelsbutareimprovedasthepassengeris
notassumedtoknowalldelaysinadvancebeforearrivingatthestation.However,themodelsstill
assumethatthepassengersknowalldelaysinadvancebutthepassengersfirstreceivethis
knowledgewhentheyarriveatthefirststation.
2
nd
generationmodelsassumethatthepassengersknowthedelaydistributionofthetrain
service(s)andincludethiswhenthepassengersconsidertheirroutesusinga1
st
generationmodel.
3
rd
generationmodelsarethemostadvancedmodels.Herethepassengerswillattempttofollowa
preplannedroutebasedonthepublishedtimetable.However,ifthisisnotpossibleorthe
passengersbecomemoredelayedthanacertainthreshold,thepassengerswillreconsidertheir
routechoicefromthatpointintimeandspace.Thisreconsiderationwillbedoneassumingfull
knowledgeofall(future)delays.

Table1belowcomparesthedifferentpassengerdelaycalculationmethodsandmodelswithrespectto
theirmaincharacteristics.

1
st
to3
rd
generationpassengerdelaymodelsrequirethesamedataandthesameworkefforttorunbut
since3
rd
generationpassengerdelaymodelsgivethemostrealisticresults(Landex2008)itis
recommendedthat3
rd
generationmodelsareused.3
rd
generationmodelsrequiremoredatathan0
th

generationmodelsbuttheyaremorepreciseandcanbeusedtoperformmorekindofanalyses.For
instance3
rd
generationmodelscanincludetransfersinthecalculationssothatthetotaltraveltimefrom
thestartstationofthepassengerstotheendstationcanbeexamined.Thismeansthatthe3
rd
generation
passengerdelaymodelscanexaminetheentirenetwork(seee.g.(LandexandNielsen2006a)andnotonly
theconsequencesofthepassengersonasinglerailwaylineorapartofthenetworkasthe0
th
generation
modelsdo(includingtheservicefrequency,traveltimeandcombinedapproaches).

TrafikdagepAalborgUniversitet2012 ISSN16039696 8
Table1:Comparisonofmethodstocalculatepassengerdelays(Landex2008).

Passengersinthe1
st
to3
rd
generationpassengerdelaymodelscanchangetheirroutechoiceincasethey
becomedelayedmorethanaspecifiedamountoftime(Landex2008;Landex,Nielsen2006b).However,in
the3
rd
generationmodelsthepassengersareassumedtoreconsidertheirroutechoiceatthepointintime
andspacewhenthespecifiedamountofdelayoccursandnotalreadybeforethetripstartsasin1
st
to2
nd

generationmodels(Landex2008).Becausepassengersdonotknowthedelaysinadvance,thepassengers
routechoicein3
rd
generationpassengerdelaymodelsarefairlyaccurate.Thismakesitpossibletoestimate
theamountofpassengersintrains,andtherebyiftheseatingcapacityissufficientbothfortheplanned
andrealizedtimetable,seeexampleinFigure6.Furthermore,capacityconstrainscanbeincorporatedin
themodelsothatitispossibletoexamineadditionaldelaysforpassengersincasetheycannotboarda
completelyfulltrainandtherebyexperiencereducedreliabilityoftherailwayoperation.

Calculatingpassengerdelaysoftheperformedoperationisofinteresttoevaluatethereliabilityofthe
systemandtoidentifyaspectsorroutinesthatcouldbeimproved.However,bycombiningthepassenger
delaymodelwitharailsimulationmodelitispossibletopredictorestimatethefuturepassengerdelays.
Thiscanbeusedtoevaluatechangesintheinfrastructureand/orinthetimetablesasearlyasinthe
planningphase.


T
r
a
i
n

d
e
l
a
y
s

(
0
t
h

g
e
n
e
r
a
t
i
o
n
)

C
r
o
s
s

s
e
c
t
i
o
n

d
e
l
a
y
s

(
0
t
h

g
e
n
e
r
a
t
i
o
n
)

C
o
u
n
t
i
n
g

t
r
a
i
n

d
e
l
a
y
s

(
0
t
h

g
e
n
e
r
a
t
i
o
n
)

O
p
t
i
m
a
l

r
o
u
t
e

c
h
o
i
c
e

m
o
d
e
l

(
1
s
t

g
e
n
e
r
a
t
i
o
n
)

1


g
e
n
e
r
a
t
i
o
n

m
o
d
e
l

P
a
s
s
e
n
g
e
r

d
e
l
a
y

m
o
d
e
l

(
2
n
d

g
e
n
e
r
a
t
i
o
n
)

P
a
s
s
e
n
g
e
r

d
e
l
a
y

m
o
d
e
l

(
3
r
d

g
e
n
e
r
a
t
i
o
n
)

Considerations of
passenger delays
No Partly Partly Partly Partly Yes Yes
Complexity of the
method
Very
simple
Low Low Medium Medium High High
Needs of
information on
passenger
demand
No
Average
alighting
passengers
Counted
passengers
OD matrix
OD
matrix
OD
matrix
OD matrix
Passengers may
predict delays in
the future (full
information is
assumed)
No No No Yes Yes Partly
Can be
incorporated
Passengers may
arrive before time
if a better con-
nection emerges
No No No Yes Yes Yes Yes
Accuracy Very low Quite low Fairly low Low Medium Medium High
Bias
Mostly
under-
estimates
delays
Will quite
often under-
estimate
delays
Will fairly
often
under-
estimate
delays
Large
under-
estimation
of delays
Under-
estimates
delays
No sys-
tematic
bias
No system-
atic bias
TrafikdagepAalborgUniversitet2012 ISSN16039696 9

Figure6:Seatingcapacityandamountofpassengersforplanedandrealizedtimetable.Basedon(Seestetal.2005).

Calculationofpassengerdelaysrequiresresultdatafromthesimulationofboththeplannedandallthe
realized/simulatedtimetablesforallarrivalsanddepartures.Thesedatacanbederivedfromtherailway
simulationsoftwareandusedinthepassengerdelaymodel.

TheworkflowofcalculatingthepassengerdelaysisshowninFigure7.Here,thesimulationofoperation,
exporttothepassengerdelaymodel,andcalculationofpassengerdelayssimulatestheimpactsfromone
toseveraldaysofoperation.Tocalibratethemodelandtoobtainadelaydistribution,itisnecessaryto
repeatthisstepanumberoftimesbeforetheevaluationisdone.

Figure7:Workflowofsimulatingdisturbancesandmodellingexpectedtrainpassengerdelays.Basedon(Landex,Nielsen2006b).

(Landex,Nielsen2006a)hasshownthepossibilityofpredictingthefuturepassengerdelaysbysimulation
fortheentireCopenhagensuburbanrailwaynetwork.Asexpected,theresultsshowedthatthepunctuality
oftrainsishigherthanthepunctualityofpassengers,cf.Figure8.

Train route B from Hje Taastrup to Holte, Train number 60126


0
50
100
150
200
250
300
350
Actual (delayed) timetable Plannedtimetable Seatingcapacity
A
m
o
u
n
t
o
f

p
a
s
s
e
n
g
e
r
s
Coding of infrastructure
Timetabling
Rules of operation
Set up of delay distribution
Simulation of operation Export to passenger delay model
Calculation of passenger delays
Evaluation
Calibration
Joi nt eval uation
Passenger del ays
Tradi ti onal si mul ation
Calibration
TrafikdagepAalborgUniversitet2012 ISSN16039696 10

Figure8:Punctualityoftrainsandpassengersatallstationsforanaverageday.Thefigureshowstheaccumulateddistributionof
arrivals,i.e.howmanypassengersarrivewithlessthanxsecondsofdelayse.g.,20%arrivebeforeplanned,60%withlessthan
60secondsofdelay(includingthe20%arrivingbeforeplanned)inthe3rdgenerationmodel.Basedon(Landex,Nielsen2006a).

Incomparison,thetraditionalwayofcalculatingpassengerpunctuality(multiplyingthedelayofthetrain
withtheexpectednumberofpassengersalightingthetrain)resulted,asexpected,inhigherpassenger
punctualitythanwhencalculatedbythepassengerdelaymodel.However,traindelaysdonotnecessarily
causepassengerdelays.Forthesimulatednetworkthe1
st
and3
rd
generationmodelscalculatethatabout
20%ofpassengersarriveearlierthanplanned,cf.Figure8.

FromFigure8itisseenthatthe1
st
generationbenchmarkmodel,asexpected,resultsinbetterpunctuality
thanthemorerealistic3
rd
generationmodel.However,thedifferenceinthemethodsis(inthiscase)small.

Apassengerdelaymodelcanalsobeusedforevaluating(andranking)infrastructureimprovements.The
benefitsforpassengersintermsoftraveltimeanddelayscanbeestimatedandcomparedwiththe
constructioncostsin,e.g.,acostbenefitanalysis.Furthermore,differenttimetablealternativescanbe
evaluatedandcomparedintheprocessofdevelopingthebestpossibletimetableforpassengers.Inthis
way,itcanbesaidthatcalculationofpassengerdelaysisofimportanceforthepassengers,thetrain
operatingcompany(e.g.,asatooltoimprovethetimetablesforthepassengersandtherebyattractmore
passengers)andtheinfrastructuremanager/planningauthoritytoprioritizeinfrastructure/maintenance
projects.

Althoughpassengerdelayscanbeinterestingtotheinfrastructuremanagerandthetrainoperating
company,thesedelaysmightnotbeusedindailyoperation.Thisisbecausetheinfrastructuremanager,
andpossiblyalsothetrainoperatingcompany,mayfocusmoreonreducingthetraindelaysthanthe
passengerdelays.Thisisbecausetraindelaysisaneasiermeasurementtodecideonandfewertraindelays
reducestheriskofconsecutivedelaysandrequireslessreschedulingofcrewandrollingstock.
Furthermore,boththeinfrastructuremanagerandthetrainoperatingcompanymaybemeasuredonthe
punctualityoftrains,ratherthanthatofpassengers,andthecompaniesmighthavecontracts(e.g.,with
theMinistryofTransportorwitheachother)resultinginabonusifthetrainpunctualityisaboveacertain
level.

0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
E
a
r
l
i
e
r

t
h
a
n

p
l
a
n
n
e
d
A
t

t
i
m
e
3
0
s

d
e
l
a
y
e
d
6
0
s

d
e
l
a
y
e
d
9
0
s

d
e
l
a
y
e
d
1
2
0
s

d
e
l
a
y
e
d
1
5
0
s

d
e
l
a
y
e
d
1
8
0
s

d
e
l
a
y
e
d
2
1
0
s

d
e
l
a
y
e
d
2
4
0
s

d
e
l
a
y
e
d
2
7
0
s

d
e
l
a
y
e
d
3
0
0
s

d
e
l
a
y
e
d
3
3
0
s

d
e
l
a
y
e
d
3
6
0
s

d
e
l
a
y
e
d
3
9
0
s

d
e
l
a
y
e
d
4
2
0
s

d
e
l
a
y
e
d
4
5
0
s

d
e
l
a
y
e
d
4
8
0
s

d
e
l
a
y
e
d
5
1
0
s

d
e
l
a
y
e
d
5
4
0
s

d
e
l
a
y
e
d
5
7
0
s

d
e
l
a
y
e
d
6
0
0
s

d
e
l
a
y
e
d
1
5
0
0
s

d
e
l
a
y
e
d
3
0
0
0
s

d
e
l
a
y
e
d
Cross section delays
Train delays
1st generation model
3rd generation model
TrafikdagepAalborgUniversitet2012 ISSN16039696 11
4 Reliability measures for highfrequent operation
Forhighfrequenttrainservices(likemetrosystems),withheadwaytimesof1to5minutes,reliability
methodsbasedontraindelaysarenotsufficient.Thisisbecausethepassengerswillnotrealizethatthe
trainsaredelayedifalltrainsaredelayedthesameamountoftime.Infact,thepunctualityfortimetables
withasimplestructureandhighfrequency(e.g.metroservices),thepunctualityisnotsoimportant(Weits
2000).Instead,passengerdelayscanbeused,butthisapproachiscomprehensive,andtheapproachonly
measuresthetrainoperationindirectly.Thereasonforthisisthatitispossiblethatapassengerispunctual
orevenaheadoftimewhenusingapreviouslydelayedtraininsteadoftheoriginallyscheduledone(Martin
2008).Therefore,differentapproachestomakeupthereliabilityforhighfrequentoperationareneeded.In
thisarticle,threeapproachesforhighfrequentoperationaredescribed.Allmethodscanbebasedon
eitherrealizedoperationsdataorsimulateddata.

4.1 Service frequency


Thefrequencyoftheoperationisaqualitycriterionforpublicpassengertransport(EuropeanCommittee
forStandardization2002).Therefore,onewayofexaminingthereliabilityoftheoperationforhigh
frequentrailwayoperationistoexaminetheservicefrequency(Kaasetal.2008).Thiscanbedoneby
examiningtheplannednumberoftraindeparturesandtherealized(orsimulated)amountoftrain
departurespertimeinterval.Iftheamountoftrainspertimeintervalisstable,itindicatesthatthetrains
arriveatthestationin(moreorless)thesametimeintervals,whichindicatesthattheoperationisstable.

Theservicefrequencyapproachisastraightforwardmethodtoevaluatethefrequencyandthereby
implicitlythereliabilityoftheoperationofhighfrequencyrailwaysystems.However,iftheexamined
timeintervalsaretoolongthereisariskofoverlookingvariationinthefrequencyandtherebythe
reliability.Therefore,thetimeintervalsasaminimumshouldbelongenoughtocontain34trains(Landex
2009).

4.2 Travel time


TheServicefrequencyapproachonlyconsiderstheheadwaytimesbetweenthetrainsandnotextended
traveltimes.Thetraveltimesbetweenthestationscanbelongerthanplannedalthoughthepromised
servicefrequencyiskept.Thiscanbethecase(fore.g.driverlessmetrosystems)ifthetrainoperating
companyoperatesmoretrainsthanneededtomaintainthepromisedfrequency.Furthermore,thetravel
timeisanimportantqualitycriterionforpublicpassengertransport(EuropeanCommitteefor
Standardization2002).Therefore,ameasurementtakingthetraveltimeintoaccountcanbewanted.

Toevaluateifthetrainoperatingcompanykeepsthepromisedquality,thetraveltimesbetweenthe
stationscanbeexamined.Forthisitcanbedecidedifthereshouldbeanacceptcriteria(orthreshold)for
whenatrainisdelayed.Fortimetableswithoutsupplements,thethresholdisnecessarysincemanytrains
elsewouldberegisteredasdelayed.Totakethepassengersexperienceintoaccount,thearticlesuggests
thattheacceptcriteriafortimetables(withandwithoutsupplements)forhighfrequentoperationshould
bethepublishedtraveltimes.

Thetraveltimemeasurementcaneitherbeexaminedbetweentwosuccessivestationsoraccumulated
traveltimesbetweenanumberofstations.Somedelaysmaynotbedetectedifitistheaccumulatedtravel
timesthatareexamined.However,sincemanypassengerstravelfurtherthanonestation,passengersmay
(duetothedifferencebetweentheminimumandthepublishedtraveltime)arriveontimealthoughthere
isa(smaller)delay.

Thetraveltimeapproachreflectstheexperiencesofthepassengersregardingtraveltimedelayinthe
vehiclebutnotdelaysduetoanunevenfrequency,cancellationsandnopossibilitytoboardatraindueto
lackofcapacity.Therefore,thetraveltimeapproachcannotcalculatethetotaldelayforpassengers.
However,theapproachissimpleandonlytherealizedtimetableisneededfortheanalyses.

TrafikdagepAalborgUniversitet2012 ISSN16039696 12
4.3 Combination of the service frequency and travel time approaches
Theservicefrequencyapproachandthetraveltimeapproacharebothsimpleandstraightforwardbutthe
methodscoverdifferentcriteria,cf.Table2.Therefore,thetwoapproachescanbecombinedtogivea
moreaccuratepictureoftheoperationquality.

Table2:Combiningtheservicefrequencyandtraveltimeapproaches(Basedon(Landex2009)).


Table2showsthatthecombinedapproachcoverstheunionoftheservicefrequencyandthetraveltime
approaches.Inthiswaythecombinedapproachcoversfrequency,reliabilityandinvehicletimeina
straightforwardwaybasedontherealizedtimetable.Furthermore,thecombinedapproachcangivea
guesstimateofthetotaltraveltimeifthetravelpatternofthepassengersisknown.Tohaveamoreprecise
estimateofthetotaltraveltimeandthedelaysofpassengersandtoincludecapacityrestrictionsmore
advancedmethodshavetobeused.

AllapproachesinTable2takethepassengersexperienceintoaccountbutindifferentways.Theservice
frequencyapproachevaluatesthenumberofthedepartingtrainsatthestationsincertaintimeintervals
butdoesnottakethetraveltimeintoaccount.Bycombiningtheservicefrequencyandtraveltime
approachesitispossibletoevaluateboththefrequencyandthetraveltimeinthetrains.Inthiswayitis
possiblegivearoughestimateofthetotaltraveltimeofthepassengersiftheaveragetravellengthofthe
passengersisknown.

5 Improving future reliability


Knowinghowtomeasurereliabilityandbeingabletosimulatedelaysandpassengerdelaysoffuture
timetables,itispossibletoanalyzethefuturereliability.Thisisbecausedelayscanbetakenintoaccountin
theplanningprocess.Timetablescanbesimulatedforexpecteddelaysforbothtrainsandpassengers
andthebestpossibletimetablecanbechosen.Thisapproachcanalsobeusedwhenplanningtimetables
forcontingencyoperationsothebesttimetablecanbeusedincaseofdisturbedoperation.

Analyzingdifferentfuturescenarios,itisalsopossibletodecidewhetheratrainshouldwaitforadelayed
connectingtrain.Notwaitingwillresultinadditionaldelaysforthepassengersinthedelayedtrain
transferringtotheothertrainresultinginreducedreliabilityforthetransferringpassengers.However,the
passengersinthetrainthatdoesnotwaitwillnotbecomedelayed,andhenceexperienceimproved
reliability.Incasethetrainhassufficienttimesupplementandthereforecancatchupthedelaybefore
thenextmajorstation,mostpassengersinthewaitingtrainwillnotexperiencereducedreliabilityalthough
thetrainwaits.Insuchacase,itmayimprovetheoverallreliabilityoftherailwaysystemthatthetrain
waits.

Whensimulatingfuturetimetablesandanalyzingdelaysfortrainsandpassengers,itispossibletooptimize
thetransfertimebasedontheexpecteddelaydistributions.Inthisway,thetransfertimecanbeadjusted
sothattheriskofmissingthecorrespondingtrainisreducedwithouthavingatoohightransfertime.This
approachimprovesthereliabilityoftherailwayoperationseenfromthepassengersperspective.Inthe
longerterm,adecisionsupportsystemcanbeusedinthetrafficcontrolcentretodecideifacorresponding
trainshouldwaitforadelayedtrain.
Servicefrequency Traveltime Combined
Frequency Yes No Yes
Invehicletime No Yes Yes
Totaltraveltime No No Roughestimate
Capacityrestrictions No No No
Complexity Low Low Low
Requireddata Realizedtimetable Realizedtimetable Realizedtimetable
Regularity Yes No Yes
TrafikdagepAalborgUniversitet2012 ISSN16039696 13

Ifatimetablehasahighamountoftimetablesupplement,itiseasiertoachievehighreliabilityinthe
operation.However,inthecaseofnodisturbancesintheoperation,thepassengerswillspendlongertime
travellingthannecessary:thepassengersexperienceextendedtraveltime.Fromthepassengers
viewpoint,anextendedtraveltimeisbetterthanadelayasitispossibletoplaninrelationtoascheduled
delay.Therefore,socioeconomiccalculationsgenerallyhaveahighervalueforunplanneddelaysthanfor
thetraveltime(includingscheduleddelays)
9
.

Beingabletosimulatepassengerdelaysoffuturetimetables,andknowingthesocioeconomicvaluesof
timeforscheduledandunscheduleddelays,itispossibletoestimatethebestleveloftimetable
supplementinthetimetable.Figure9showstheidealizedsocioeconomicutility(withagivensetof
parameters)dependingonthetimetablesupplement,andtherebyimplicitlytherunningtimeand
(expected)delayedtime.

Figure9:Socioeconomicutilityofagiventimetable(foranidealizedsituation)dependingontheamountoftimetable
supplement(Landex2008).

ThismethodhasbeenusedforpreliminaryanalysesoftheDanishrailwaylinefromRingstedtoNykbing
Falster.Here,Thorhauge(2010)demonstratedthattheoptimumtimetablesupplementsfordifferent
timetablescenarioswere78%whilethecurrenttimetablesupplementwas12.612.7%onaverage.

6 Conclusions
Thisarticledefinesreliabilityasthecontinuityofcorrectservicethatcanbedividedintothetwoclosely
relatedconceptsregularityandpunctuality.Thisisfollowedbyimprovedmethodstomeasurereliabilityof
railwayoperationforbothtrainsandpassengersforconventionalaswellashighfrequentoperation.These
improvedmethodsareenabledbythedevelopmentofsimulationmodelsandthepossibilityofcombining
thesesimulationmodelswithroutechoicemodels.Toachievethebestresultswiththenewmore
sophisticatedmethods,moredataandmodelcalibrationisrequired.

Usingthemeasurementsonrealizedaswellassimulatedrailwayoperationcanlocatereliabilityproblems
tobedealtwithandtherebyimprovetheservice.Thiscanbedonebothfornormalandcontingency
operation.Thereliabilitycanbeimprovedbyincludingmoresupplementsinthetimetableincreasingthe
traveltime.However,thereisatradeoffbetweenreliabilityandthetraveltime.Usingsimulationandthe
presentedreliabilitymeasures,theoptimaltimetablesupplementscanbefoundbasedonasocio
economicanalysis.

9
In Denmark, an unplanned delay is valued twice as much as travel time (including scheduled delays) (Danish Ministry of Transport 2006).
Socioeconomic optimum
of timetable supplement
Timetable supplement [minutes]
S
o
c
i
o
e
c
o
n
o
m
i
c

u
t
i
l
i
t
y

[

]
TrafikdagepAalborgUniversitet2012 ISSN16039696 14
7 References
Avizienis,A.,Laprie,J.&Randell,B.2001,FundamentalConceptsofDependability,Computerand
InformationScience,vol.1145,pp.712.

Barber,F.,Abril,M.,Salido,M..,Ingolotti,L.P.,Tormos,P.&Lova,A.2007,SurveyofautomatedSystems
forRailwayManagement,DepartmentofComputerSystemsandComputation,TechnicalUniversityof
Valencia.

Berger,A.,Hoffmann,R.,Lorenz,U.&Stiller,S.2011,Onlinerailwaydelaymanagement:Hardness,
simulationandcomputation,Simulation,88(7),pp.616629.

Bush,R.2007,DoesEveryTripNeedtoBeOnTime?MultimodalSchedulingPerformanceParameterswith
anApplicationtoAmtrakServiceinNorthCarolina,inProceedingsof86
th
AnnualMeetingatTransport
ResearchBoard,TransportationResearchBoard,WashingtonD.C.,USA.

Daamen,W.,Goverde,R.M.P.&Hansen,I.A.2007,NondiscriminatoryAutomaticandDistinctRegistration
ofPrimaryandSecondaryTrainDelays,inProceedingsofthe2
nd
InternationalSeminaronRailway
OperationsModellingandAnalysis,eds.I.A.Hansen,A.Radtke,J.P.Pachl&E.Wendler,International
AssociationofRailwayOperationsResearch,Hannover,Germany.

DanishMinistryofTransport.2006,Manualconcerningsocioeconomicanalysisappliedmethodologyin
thetransportsector(Ngletalskatalogtilbrugforsamfundskonomiskeanalyserptransportomrdet),
4
th
edn,DanishMinistryofTransport,Denmark.(inDanish)

DAriano,A.2008,ImprovingRealTimTrainDispatching:Models,AlgorithmsandApplications,Delft
UniversityofTechnology.

EuropeanCommitteeforStandardization2002,TransportLogisticsandservicesPublicpassenger
transportServicequalitydefinition,targetingandmeasurement(EN13816),DanishStandards
Association,Charlottenlund,Denmark.

deFabris,S.,Longo,G.&Medeossi,G.2008,Automatedanalysisoftraineventrecorderdatatoimprove
microsimulationmodels,inProceedingsofthe11
th
InternationalConferenceonComputersinRailways,
eds.J.Allan,E.Arias,C.A.Brebbia,etal,WITpress,pp575.

JespersenGroth,J.,Potthoff,D.,Clausen,J.,Huisman,D.,Kroon,L.G.,Maroti,G.&Nielsen,M.N.2007,
DisruptionManagementinPassengerRailwayTransportation,InformaticsandMathematicalModelling,
TechnicalUniversityofDenmark.

Hoyle,D.1994,ISO9000QualitySystemsHandbook,2
nd
edn,ButterworthHeinemannLtd,Oxford,Great
Britain.

Jacobs,J.2008,Rescheduling,inRailwayTimetable&Traffic,eds.I.A.Hansen&J.Pachl,1
st
edn,Eurail
Press,pp182191.

Kaas,A.H.&Jacobsen,E.M.2008,AnalysingtheMetroCityringinCopenhagen,Proceedingsofthe11
th

InternationalConferenceonComputersinRailways,eds.J.Allan,E.Arias,C.A.Brebbia,etal,WITpress,pp.
45.

Kaas,A.H.,Jacobsen,E.M.&Lund,K.H.2008,TheCityringtrafficanalyses(Cityringen,trafikaleanalyser),
inProceedingsoftheAnnualTransportConferenceatAalborgUniversity,inDanish.

TrafikdagepAalborgUniversitet2012 ISSN16039696 15
Kaas,A.H.2000,Punctualitymodelforrailways,inProceedingsofthe7
th
InternationalConferenceon
ComputersinRailways,eds.J.Allan,R.J.Hill,C.A.Brebbia,G.Sciutto,S.Sone,WITpress,pp.853.

Kaas,A.H.1998,Methodstocalculatecapacityofrailways(Metodertilberegningafjernbanekapacitet),
TechnicalUniversityofDenmark,inDanish.

Koutsopoulos,H.N.&Wang,Z.2007,SimulationofUrbanRailOperations:ApplicationFramework,
JournaloftheTransportationResearchBoard,vol.2006,pp.8491.

Landex,A.2009ExaminationofOperationQualityforHighfrequentRailwayOperation,inProceedingsof
the3
rd
InternationalSeminaronRailwayOperationsResearch,IAROR,Switzerland.

Landex,A.2008Methodstoestimaterailwaycapacityandpassengerdelays,TechnicalUniversityof
Denmark,DepartmentofTransport.

Landex,A.,Kaas,A.H.&Nielsen,O.A.2007,PunctualityofRailways(Opgrelseafregularitetp
jernbaner),ProceedingsoftheAnnualTransportConferenceatAalborgUniversity,inDanish.

Landex,A.&Nielsen,O.A.2006,"Modellingexpectedtrainpassengerdelaysonlargescalerailway
networks",Proceedingsofthe7
th
WorldCongressonRailwayResearch,UIC,Canada.

Landex,A.&Nielsen,O.A.2006,Simulationofpassengerdelaysinrailwaynetworks(Simuleringaf
passagerforsinkelserpjernbaner),inProceedingsoftheAnnualTransportConferenceatAalborg
University,eds.L.G.Hansen,L.D.Nielsen,O.A.Nielsen&T.Hels,inDanish.

Martin,U.2008,PerformanceEvaluationinRailwayTimetable&Traffic,eds.I.A.Hansen&J.Pachl,1
st

edn,EurailPress,pp.192208.

NEA2003,BOBRailwayCaseBenchmarkingPassengerTransportinRailways,NEATransportresearchand
training,Rijswijk,TheNetherlands.

Nash,A.&Huerlimann,D.2004,RailroadsimulationusingOpenTrack,inProceedingsofthe9
th

InternationalConferenceonComputersinRailways,eds.J.Allan,C.A.Brebbia,R.J.Hill,G.Sciutto&S.Sone,
WITpress,pp45.

NederlanseSpoorwegen2001,DeKoppeling,nr.1619,NederlandseSpoorwegen,Utrecht,The
Netherlands,inDutch.

Nie,L.&Hansen,I.A.2005,Systemanalysisoftrainoperationsandtrackoccupancyatrailwaystations,
EuropeanJournalofTransportandInfrastructureResearch,vol.5,no.1,pp.3154.

Olsson,N.O.E.&Haugland,H.2004,Influencingfactorsontrainpunctualityresultsfromsome
Norwegianstudies,TransportPolicy,vol.11,no.4,pp.387397.

Oort,N.van&Nes,R.van2010,ImpactofRailTerminalDesignonTransitServiceReliability,Journalof
TransportationResearchBoard,vol.2146,pp.109118.

Oort,N.van&Nes,R.van.2005,ServiceRegularityAnalysisforUrbanTransitNetworkDesign,in
Proceedingsofthe10
th
InternationalConferenceonComputerAidedSchedulingofPublicTransport.

Radtke,A.&Bendfeldt,J.2001,HandlingofrailwayoperationproblemswithRailSys,inProceedingsof
the5
th
WorldCongressonRailResearch,Cologne,Germany
TrafikdagepAalborgUniversitet2012 ISSN16039696 16

Seest,E.,Nielsen,O.A.&Frederiksen,R.D.2005,CalculatingpassengerpunctualityintheCopenhagen
suburbannetwork(OpgrelseafpassagerregularitetiStog),inProceedingsoftheAnnualTransport
ConferenceatAalborgUniversity,inDanish.

Siefer,T.2008,Simulation,inRailwayTimetable&Traffic,eds.I.A.Hansen&J.Pachl,1
st
edn,EurailPress,
Hamburg,Germany,pp.155169.

Siefer,T.&Radtke,A.2005,RailwaySimulationKeyforBetterOperationandOptimalUseof
Infrastructure,inProceedingsofthe1
st
InternationalSeminaronRailwayOperationsModellingand
Analysis,eds.I.A.Hansen,F.M.Dekking,R.M.P.Goverde,B.Hindergott&L.E.Meester,International
AssociationofRailwayOperationsResearch,Delft,TheNetherlands.

Thorhauge,M.2010,Optimizationoftimetablesupplementfromapassengerbasedsocioeconomicpoint
ofview,inProceedingsoftheAnnualTransportConferenceatAalborgUniversity,Trafikdage,Aalborg,
Denmark.

Toledo,T.,Koutsopoulos,H.N.2004,Statisticalvalidationoftrafficsimulationmodels,Journalofthe
TransportationResearchBoard,vol.1876,no.29,pp.142150.

Tromp,J.P.M.2004,Validationofatrainsimulationmodelwithtraindetectiondata,inProceedingsof
the9
th
InternationalConferenceonComputersinRailways,eds.J.Allan,C.A.Brebbia,R.J.Hill,G.Sciutto&
S.Sone,WITpress,pp.583.

Vromans,M.J.C.M.,Dekker,R.Kroon,L.G.2006,Reliabilityandheterogeneityofrailwayservices,
EuropeanJournalofOperationalResearch,vol.172,no.2,pp.647665.

Vromans,M.J.C.M.2005,ReliabilityofRailwaySystems,NetherlandsTRAILResearchSchool.

Weits,E.A.G.2000,RailwayCapacityandTimetableComplexity,inProceedingsofthe7
th
International
WorkshoponProjectManagementandScheduling,Osnabrck,Germany.

Yuan,J.2006,StochasticModellingofTrainDelaysandDelayPropagationinStations,NetherlandsTRAIL
ResearchSchool.

You might also like