Railway Reliability, Analytical Models Max-plus algebra is an analytical approach for evaluating

timetables on robustness. Some key characteristics, like minimal cycle times, are easily
calculated with max-plus algebra (Subiono, 2000, Goverde and Soto y Koelemeijer, 2000, Van
Egmond, 1999, De Kort, 2000). PETER, based on max-plus algebra, is a performance evaluator
for timetables (Soto y Koelemeijer et al., 2000, Goverde and Odijk, 2002). Current maxplus
research in the field of railways focuses on the inclusion of stochastic disturbances in the
models. Hansen (2000) uses both queuing theory and max-plus algebra to study the capacity
and stability of train movements, but only in stations. Another probabilistic approach by De Kort
et al. (2003) uses a max-plus model to assess the capacity of the railway infrastructure. This
probabilistic capacity assessment is based on unspecific timetables. This means that train lines
are known, but there are no actual departure and arrival times available. For a more detailed
description of max-plus algebra and its possibilities, see Section 3.2. Weigand (1981) develops a
model that is able to evaluate delay propagation of exponential disturbances throughout a
railway network. He also shows how to determine the minimal amount of running time
supplements that is necessary to achieve a certain average arrival delay. Wakob (1985)
introduces a queueing model to assess the capacity for a given subnetwork. His method is
known as Wakob’s razor, and is based on random train arrivals without a timetable.
Schwanh¨außer (1994) describes the use of queueing models for determining the expected
value of anticipated delays and other timetable characteristics. These queueing models are
based on train frequencies and running times only, not on timetables. Therefore they are better
suited for a capacity analysis of the infrastructure than for timetable evaluation, let alone for
timetable optimization. Huisman and Boucherie (2001) and Huisman et al. (2002) developed a
stochastic analytical waiting time model for analyzing delays at a double track section. Huisman
(2005) is also able to apply these queueing models to scheduled timetables. With his models he
also analyzes the consequences of several timetable characteristics for the delay propagation.
Higgins et al. (1995) come up with a model to quantify so called risk delays for a single track line.
Higgins and Kozan (1998) also developed an analytical model to quantify the expected delays of
individual passenger trains in an urban rail network. Carey and Kwieci´nski (1995) mainly focus
on recovery times in their stochastic analysis. Carey (1999) also uses heuristic measures for
timetable reliability and includes behavioral response (Carey, 1998).