Increasing the precision in timetable planning is a key success factor for all
infrastructure managers, since it allows us to minimize delay propagation
without reducing usable capacity. Since most running time calculation models
are based on standard and deterministic parameters an imprecision is implicitly
included, which has to be compensated by running time supplements.
At the same time, GPS or even more precise trackings are continuously stored
in the event recorders of most European trains. Unfortunately, this large amount
of data is normally stored but not used except for failure and maintenance
management.
To consider real running time variability in running time calculation, an
approach has been developed, which allows us to calibrate a performance factor
for each motion phase.
Given the standard motion equation of a train, and a mesoscopic model of the
line, the tool uses a simulated annealing optimisation algorithm to find the best
regression between calculated and measured instant speed. To increase precision,
the motion is divided into four phases: acceleration, braking at stops, braking for
speed reductions/signals and cruising. By performing the procedure over a
number of train runnings, a distribution of each performance parameter is
obtained. Once the infrastructure model is defined and the trackings are
imported, the procedure is completely automated.
The approach can be used in both stochastic simulation models and as a basis
for advanced timetable planning tools, where stochastic instead of deterministic
running times are used. The tool has been tested in the north-eastern part of Italy
as input for both running time calculation and microscopic simulation.
