The influence of buffer time distributions in delay propagation modelling of railway networks. Issue 3 (December 2018)

• Record Type: Journal Article
• Authors: Zieger, Stephan; Weik, Norman; Nießen, Nils
• Is Part Of: Journal of rail transport planning & management. Volume 8:Issue 3/4(2018)
• Page Start: 220
• Page End: 232
• Publication Date: 2018-12
• Subjects: Buffer times -- Delays -- Stochastic models -- Capacity -- Railways -- Long term planning; Railroads -- Periodicals; Railroads -- Planning -- Periodicals; Railroads -- Management -- Periodicals; Railroads; Railroads -- Management; Railroads -- Planning; Periodicals; 385.068
• DOI: 10.1016/j.jrtpm.2018.09.001
• Languages: English
• ISSNs: 2210-9706

Abstract

Abstract: Buffer times are an important factor in railway timetable design preventing the propagation of delays and ensuring timetable robustness. Determining the required amount of buffer times, such that a certain level of service quality is achieved, falls within the responsibility of railway capacity analysis. This is why capacity analysis is intrinsically linked to delay propagation modelling. Currently, delay propagation modelling in this context relies on the assumption of random, exponentially distributed or deterministic buffer times. Real-world timetables tend to deviate from this behaviour, such that a more general modelling of buffer time distributions is desirable. In this paper the impact of different buffer time distributions on the build-up of knock-on delays in delay propagation modelling is analysed using a Monte-Carlo simulation approach. It is shown that the choice of distribution has a significant impact on performance metrics. In a sensitivity analysis line capacity is observed to vary by as much as 17 % as a function of the underlying buffer time statistics in the investigated scenarios. Highlights: The role of buffer time distributions in delay propagation modelling is investigated. The choice of buffer time distribution is shown to be crucial for delay build-up. Rail capacity is found to vary up to 17% depending on buffer time statistics. In a sensitivity analysis the effects of input parameters on capacity are analysed.