A Digital Railway System, and what it means for operators and users - PowerPoint PPT Presentation

A Digital Railway System, and what it means for operators and users Risk Seminar Francesco Corman francesco.corman@ivt.baug.ethz.ch F. Corman | | 23.10.2019 1

Overall concept of this slides Risk = probability x intensity x exposure x vulnerability Risk/uncertainty with smaller probability, larger exposure-vulnerability Digitalization enables accurate quantification/ feedback/ Risk/uncertainty control/ with high probability, shift small exposure-vulnerability F. Corman | | 23.10.2019 2

Challenges in railway systems  Infrastructure capacity is money intensive, inflexible, slow to build, in large steps  Operations tightly constrained and interrelated , with small operative margins  Variability in operations causes delay propagation to traffic, impacting passengers  Increasing traffic worsens performance, amplifying delay propagation  Increasing quality expectations, passengers, competition ; economic efficiency F. Corman | | 23.10.2019 3

Railway network currently used extensively and with high quality Punctuality = Quality [Dutch Railways NS, 2018] Utilization rate [train km per track km] F. Corman | | 23.10.2019 4

Delays in Switzerland exist (Graffagnino et al 2012)  Instability due to many controllable/incontrollable internal/external factors F. Corman | | 23.10.2019 5

Massive growth expected Public Transport/Railway 51% growth 2010  2040 vulnerability to instability In 2040, compared to 2010: +37% ton km +25 % passenger km +28% population increase +46% economy increase 2040 [Federal Statistic Office, BFS OFS [Federal Office for Spatial Development ARE OFDT, F. Corman | | 23.10.2019 6 Pocket Statistics 2018] Perspective 2040]

Potential from digitalization Understand Delay phenomena Fusing unreliable GPS with ? reliable track-based data Predict Determining values for future phenomena Identifying future situations ? Optimize Deciding against possible situations 0 1 1 0 0 2 3 00 0 1 @# Digital railway system F. Corman | | 23.10.2019 7

How to get there? Need to shifts risks/uncertainties How to make it possible for trains to be How to make it possible for trains to be where they should? where they should? Infrastructure Planning Where do passengers want trains to be? Where do passengers want trains to be? Timetable Planning Where should all trains be? Where should all trains be? Delay management Where will a train be? Where will a train be? Traffic management Where should a train be? Where should a train be? Train movement Train safety Where is a train? Where is a train? F. Corman | | 23.10.2019 8

How to get there? Need to shifts risks/uncertainties Risk = probability x intensity x exposure x vulnerability high low high low Infrastructure Planning Timetable Planning Delay management Traffic management Positioning Train movement Train safety F. Corman | | 23.10.2019 9

Understand - positioning  Impact of bad data collection: positioning of trains, GPS signaling system Position estimate, no measure Sessa, P.G., De Martinis, V., Corman, F (2019) CACAIE Fusing power data with speed Fusing power data with speed reduces position uncertainty reduces position uncertainty Position estimate, power measured in GPS blackout areas in GPS blackout areas to operationally useful levels to operationally useful levels F. Corman | | 23.10.2019 10

Systemic risks – the example of fully automated train operations  Railway has very high safety already  Control of speed and variability  less variations, increased punctuality  High precision control as enabler of reduction of margins  increased capacity  Risks of shutdown of system whenever a human-acceptable, but machine-unacceptable situation appears  Ultimately, reliability or performance could decrease (e.g. over protective advanced signaling systems) F. Corman | | 23.10.2019 11

How to get there? Need to shifts risks/uncertainties Risk = probability x intensity x exposure x vulnerability low to high low high low-medium Infrastructure Planning Timetable Planning Traffic operations Delay management Traffic management Train movement Train safety F. Corman | | 23.10.2019 12

We have little clue whether delays will increase/decrease with time  Railway operations and delays are interrelated stochastic processes, but most often considered as deterministic process % Accurate predictions 2 days, 4 days, 6 days ahead Accuracy of weather forecast 6 days in 2010, better than 2 days in 1970 [Deutscher Wetterdienst] F. Corman | | 23.10.2019 13

Predictions – impact into decisions affecting future traffic  “Things are almost constant” Corman Kecman ( 2017) TRC  Adjusting operating plan, departures, train crossings, transfers  Resource allocation – drivers, vehicles  Passengers might take wrong services, or face delays Büchel, Corman ( 2020) TRBoard  Bayesian network approaches might help not only predict, but also understand F. Corman | | 23.10.2019 14

Reassigning capacity in real time: Railway traffic control Based on a current network state , determine traffic control actions (retiming; reordering, rerouting, cancelling, … ) e.g. should the yellow train overtake the blue? ? time time } arrival time } arrival time  delay  delay which proactively reduce delays and delay propagation, in a short computation time F. Corman | | 23.10.2019 15

State of practice  potential of decision support F. Corman | | 23.10.2019 16

High probability, small vulnerability: delays Traffic Control Algorithms: Optimized Orders Alternative Graph Approach for optimization First In First Out Rule-based Keep the Timetable Order Variation in observed delays decreases Variation in observed delays decreases Corman et al ( 2011) TPT F. Corman | | 23.10.2019 17

Bringing passengers into the picture - delays  Compliance  Information availability and dissemination  Real time information on passengers  Existence of suboptimal User Equilibria Corman et al; (2016) TR part E Delaying trains instead of passengers: Delaying trains instead of passengers: 12% shorter travel time vs timetable 12% shorter travel time vs timetable F. Corman | | 23.10.2019 18

Marra, Corman 2020 TRBoard Small probability, larger vulnerability: disruptions Disruptions, Dutch network 2012 2013 2014 F. Corman 2015 | 23.10.2019 | 19 Rastatt Tunnel disruption, Amount of events classified as disruption, Dutch Network [Corman] august- October 2017-DB AG

Passengers as a key aspect in disruptions reactive information one order of reactive information one order of magnitude better than no information magnitude better than no information  Agent-based models estimate passenger evaluation of disruption  Modelling different information availability strategies  Variability in impact  Further potential to decrease Leng, Corman CASPT 2018 impact changes by managing the system Best reactive information only slightly Best reactive information only slightly F. Corman | | 23.10.2019 20 worse than proactive information worse than proactive information

How to get there? Planning for uncertainty Risk = probability x intensity x exposure x vulnerability low low-medium high low-medium Planning Infrastructure Planning Timetable Planning Delay management Traffic management Train movement Train safety F. Corman | | 23.10.2019 21

Switzerland typically investing heavily in hardware – railway infrastructure very expensive and long time to delivery  STEP Ausbau Schritte 2035 : 11 Billion CHF for new infrastructure F. Corman | | 23.10.2019 22 [Federal Office for Transport BAV-OFT, 2017]

Investing in timetable planning? In demand management?  Digitalization allows better quantification of performance of plans regarding the demand  How much risks in transport performance we can manage by just a better software?  How much risks in capacity/transport performance we can manage by changing demand patterns? E.g. reducing peak demand Amstutz 2019 F. Corman | | 23.10.2019 23

Conclusion - Overall concept of this slides Risk = probability x intensity x exposure x vulnerability Risk/uncertainty with smaller probability, larger exposure-vulnerability Digitalization enables accurate quantification/ feedback/ Risk/uncertainty control/ with high probability, shift small exposure-vulnerability F. Corman | | 23.10.2019 24

A Digital Railway System, and what it means for operators and users Risk Seminar Francesco Corman francesco.corman@ivt.baug.ethz.ch F. Corman | | 23.10.2019 25