Research question and state of the art

Punctuality and delays are important issues and indicators for railways, and they can negatively affect the demand to a high amount, as well as causing more direct costs to customers, operators and society at large. To measure and follow up delays are thus both common and important aspects of improving operations. Statistics and other related methods can then be used to better understand and model the distribution of delays, so that improvement measures can be prioritised and targeted appropriately. Increasingly, both planning and evaluation of railway operations being aided by simulation models aiming to estimate the reliability of the system. For instance, simulations are often used to analyse the consequences of proposed infrastructure changes or timetables, in terms of travel time, delays and capacity utilization. The simulations can be performed with microscopic simulation models, where everything is modelled in detail, or more general with macroscopic models, depending on the requirements of the application. An important aspect for simulations is how to handle incomplete or unknown input data, e.g. the departure times of freight trains in Sweden. They do not necessarily depart according to the timetable, but rather when they are ready and there is a free train path in the timetable. The ramifications of this are not obvious, in terms of impacts on delays and surrounding traffic, and the phenomenon is not easy to model in existing simulation software.

The aim of this study is to investigate how a microscopic simulation model (RailSys) and a macroscopic simulation model (PRISM) handle simulations with early freight train departures. Two main questions are answered in a simulation case study:

• How do the simulation results differ between simulations with no early freight trains and simulations with early freight trains? This is answered by comparing PRISM and RailSys simulation results with and without early freight train departures, respectively.
• Which simulation practice - with or without early freight trains - reflects reality best? This is answered by comparing the simulation results with empirical data.

Microsimulation excels in modelling railway infrastructure and rolling stock in detail, attaining a high degree of accuracy given a system for driving and dispatching. Several different microscopic simulation tools are available, e.g. RailSys (Bendfeldt, Mohr and Müller, 2000; Radtke and Hauptmann, 2004), LUKS (Janecek and Weymann, 2010), and OpenTrack (Nash and Huerlimann, 2004).

When larger networks are simulated, the amount of details in microscopic models results in both unacceptably long simulation times, and increased complexity for the user, who has to model the network in miniscule detail. Macroscopic models, where only the most critical infrastructure elements (such as lines and stations) and the most critical events (such as train departures and arrivals) are represented, can therefore be preferred. A macroscopic model for delay propagation in large networks was presented by Büker and Seybold (2012). Zinser et al. (2018) presented a macroscopic simulation model and performed a case study comparing it to a microscopic simulation approach for infrastructure disruptions, showing promising results for the new model. This model was subsequently named PRISM and further developed (Zinser et al., 2019), creating what is essentially a macroscopic simulation tool, but with the possibility to adjust the level of detail to enhance its range from macroscopic to microscopic simulations, depending on the required accuracy. Another adjustable model capable of micro-, macro- and mesoscopic (i.e. intermediate level of detail) simulations was introduced by Cui, Martin and Liang (2018). Efforts have also been made to integrate microscopic and macroscopic models to simplify data management while improving timetable forecasting (Huber and Wilfinger, 2006).

Method

A simulation network of the line Hallsberg – Malmö, an important Swedish freight corridor, is set up in RailSys and PRISM. To create realistic simulations, delay distributions for initial, dwell and run time delays are created from real delay data, with the dwell and run time distributions scaled down to remove secondary delays. The initial delay distributions are created in two versions: one allowing early departures for freight trains and one allowing only late or on time departures. Since PRISM allows negative delays as input, initial delay distributions with early freight trains are straightforward to use. RailSys, however, does not accept negative delays and the simulation timetable and initial delays have to be shifted first, as described by Lindfeldt and Sipilä (2014). Simulation is then performed in RailSys and PRISM with the timetable for a normal autumn Thursday in 2016.

Analysis and results

The planned analysis is to plot cumulative distributions of arrival delays for freight and passenger trains in PRISM (with and without early freight trains), RailSys (with and without early freight trains) and empirical data, perform two-sample Kolmogorov-Smirnov tests to see if they are part of the same distributions. The same is done for punctuality, i.e. RT+5 at destination, and potentially also for dwell and run time delays.

References

Bendfeldt, J. P., Mohr, U. and Müller, L. (2000) ‘RailSys, a system to plan future railway needs’, Computers in Railways VII, pp. 249–255.

Büker, T. and Seybold, B. (2012) ‘Stochastic modelling of delay propagation in large networks’, Journal of Rail Transport Planning and Management, 2(1–2), pp. 34–50. doi: 10.1016/j.jrtpm.2012.10.001.

Cui, Y., Martin, U. and Liang, J. (2018) ‘PULSim: User-Based Adaptable Simulation Tool for Railway Planning and Operations’, Journal of Advanced Transportation. doi: 10.1155/2018/7284815.

Huber, H.-P. and Wilfinger, G. (2006) ‘Integration-Enhancements for Microscopic and Macroscopic Railway Infrastructure Planning Models’, in Proceedings of 7th World Congress on Railway Research (WCRR) 2006. Montreal, Canada.

Janecek, D. and Weymann, F. (2010) ‘LUKS – Analysis of lines and junctions’, in 12th World Conference of Transport Research. Lisbon, Portugal.

Lindfeldt, A. and Sipilä, H. (2014) ‘Simulation of freight train operations with departures ahead of schedule’, WIT Transactions on the Built Environment, 135, pp. 239–249. doi: 10.2495/CR140191.

Nash, A. and Huerlimann, D. (2004) ‘Railroad simulation using OpenTrack’, Computers in Railways IX, pp. 45–54.

Radtke, A. and Hauptmann, D. (2004) ‘Automated planning of timetables in large railway networks using a microscopic data basis and railway simulation techniques’, Computers in Railways IX, pp. 615–625.

Zinser, M. et al. (2018) ‘Comparison of microscopic and macroscopic approaches to simulating the effects of infrastructure disruptions on railway networks’, in Proceedings of 7th Transport Research Arena TRA 2018, April 16-19, 2018. Vienna, Austria.

Zinser, M. et al. (2019) ‘PRISM : A Macroscopic Monte Carlo Railway Simulation’, in Proceedings of 12th World Congress on railway Research (WCRR) 2019. Tokyo, Japan.