Time: Thursday, 05/Sept/2024: 11:30am - 1:00pm Session Chair: Carsten Moldenhauer |
Location: Nordgebäude ZG 1070 Room Location at NavigaTUM |
Dimensioning of track groups in rail freight stations on the basis of an optimised waiting probability
Technische Universität Dresden, Germany
The quality-oriented and economical dimensioning of track groups (bowls) in rail freight stations is an essential task in the sustainable design of the railway network at the strategic planning level. The main task here is to determine the required number of tracks and the necessary track characteristics. In contrast to pure passenger railway stations, which today in Europe are mostly served with regular timetables and therefore experience cyclically repeating track occupancy, the performance requirements in rail freight stations are more irregular: the arrival times of the trains and the track occupancy times are often distributed rather randomly. In addition, freight train handling involves significantly more complex processes and dependencies. At present, track groups in freight stations are mainly dimensioned using analytical queueing models or simulations in which the waiting processes in front of a track group are analysed. In most cases, the waiting probability is considered as the decisive parameter. This paper describes an optimisation model that maps the track occupancy in a track group, taking into account the handling processes and complex dependencies, and minimises the number of trains waiting in front of the track group. The input data is generated using given distribution functions which anticipate expected characteristics of the later operation. Practical scenarios are used to illustrate how the approach works. In addition, a quantitative comparison is made with the methods used to date and alternative approaches for the dimensioning of track groups.
Scheduled network design for single wagon load transportation at the Swiss National Railways
SBB Cargo AG, Switzerland
The Swiss National Railways operate a single wagon load transportation system in which customers can ship demands that do not justify a unit train. Moreover, these demands are time sensitive, in particular for overnight express connections. To make this work, demands are consolidated at multiple intermediate stops, or classification yards, onto scheduled trains.
The computation of the trains and their timings, and the transportation chains of the demands along the trains, is a network design problem that we model as a fixed-charge multi-commodity-flow problem on a time-expanded network.
The first main contribution of this work is the computation of small time-expansions despite the fine granularity of the intended schedules. Our methods are, in spirit, related to the dynamic discretization discovery methods by Boland and others. But in contrast, they provide an a priori sparsification of the time expansion and do not require iterative resolves ot the optimzation model.
These small time expansions then permit an effective solution of the induced integer linear program.
The second main contribution are several postprocessing mechanisms to increase the applicability of the obtained network designs for practice.
The computed network designs have been used at SBB Cargo for tactical planning and commercial client offers. Experimental evaluations on some test instances from these practical examples is also provided.
CANCELLED: An Integer Programming Model to Assign Train Drivers to Good Positions in Basic Turni
1Technische Hochschule Wildau, Germany; 2COSMO Consult, Germany
We are reporting on a project on duty rostering of a train operating company in Germany. There, in the past, rostering took place on a purely individual basis. As the goal of the project, the train drivers should work according to a set of 20 essentially fixed basic turni. These plans took into effect from January 2024 on.
In this extended abstract, we are focusing on a specific task that had to be resolved within the transition process. In particular, the fixed turni are intended to repeat periodically, most of them after 17 weeks. In particular, for that one turnus is equipped with train drivers in a balanced way, 17 employees have to be assigned to this turnus, each of them to one of the 17 different weeks of the turnus to start. To cover each day of the week, the weeks of the turnus typically differ in their number of working days.
Since the holidays of the train drivers had been already planned in advance, depending on the starting week, during the entire year, a smaller or larger number of working days of the turnus could be erased by the individual holidays. In order to preserve as many working days as necessary, a straightforward mixed-integer linear optimization problem has been designed and solved, to decide which train driver should start to work in which week of the turnus. The solution of this optimization run has been finally applied in practice.