Simulation of rail traffic: applications with timetable construction and delay modelling
2012 (English)Licentiate thesis, comprehensive summary (Other academic)
This thesis covers both applications where simulation is used on parts of the Swedish rail networks and running time calculations for future high-speed trains with top speed improvements on existing lines. Calculations are part of a subproject within the Green Train research program (Gröna tåget). Higher speeds are possible with increased cant and cant deficiency in curves. Data for circular curve radii is used on existing lines combined with information on decided and on-going upgrades. Calculation of static speed profiles is made for a set of cant and cant deficiency values. Different train characteristics are used regarding top speed, starting acceleration and power to ton ratio. Running time calculations are made for these different train characteristics with the fictive speed profiles. In addition, different stopping patterns are applied. Results are presented together with running times for two reference train types, one with carbody tilting and one without. It is clear that carbody tilting, allowing a higher cant deficiency, is important on many of the existing lines considering achieved running times. The benefit of tilting is marginal on newly built and future lines designed with large curve radii. However, on many of the existing lines the over 20 year old reference train with carbody tilting achieves shorter running times compared to a future train without tilt but with higher top speed. The work presented here has contributed with input to other projects and applications within the research program. Simulation in RailSys is used to evaluate on-time performance for high-speed trains, between Stockholm and Göteborg in Sweden, and changes in timetable allowances and buffer times with respect to other trains. Results show that ontime performance can be improved with increased allowances or buffer times. In the case with increased buffers, other trains are pushed in the timetable with the intention of obtaining at least five minutes at critical places (e.g. conflicting train paths at stations) and as separation on line sections. On-time performance is evaluated both on aggregated (group) level and for trains individually. Some of the trains benefit significantly from the applied measures. Prior to a simulation some of the delays have to be defined. This includes dwell extensions and entry delays, i.e. extended exchange times at stations and delayed origin station departures inside or at the network border. Evaluation of observed data give insight on the performance of a real network. However, separating primary (exogenous) and secondary (knock-on) delays is not straightforward. Typically the probabilities and levels of primary delays are defined as input, thus secondary delays are created in the simulations. Although some classification of delays exist in observed data, it is not sufficient without further assumptions and preparation. A method for estimating primary running time extensions is presented and applied on a real timetable between Katrineholm and Hässleholm in Sweden. The approach consist of creating distributions based on deviations from scheduled running time. Since this represent total outcome, i.e. both primary and knock-on delays are included, the distributions are reduced by a certain percentage and applied in the simulations. Reduction is done in four steps, separately for passenger and freight trains. Root mean square error (RMSE) is used for comparing mean and standard deviation values between simulated and observed data. Results show that a reasonably good fit can be obtained. Freight services show a higher variation than passenger train evaluation groups. Some explanations for this are difficulties in capturing the variations in train weights and speeds and absence of shunting operations in the model. In reality, freight trains can also frequently depart ahead of schedule and this effect is not captured in the simulations. However, focus is mostly on passenger trains and their on-time performance. If a good enough agreement and operational behaviour is achieved for them, a lower agreement for freight trains may be accepted.
Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2012. , viii, 54 p.
Transport Systems and Logistics
IdentifiersURN: urn:nbn:se:kth:diva-97461ISBN: 978-91-85539-87-1OAI: oai:DiVA.org:kth-97461DiVA: diva2:533359
Holst, Anders, Assoc. professorEkman, Jan, PhD
Nelldal, Bo-Lennart, Professor
QC 201206112012-06-132012-06-132012-06-13Bibliographically approved
List of papers