Long-term capacity planning of railway infrastructure : a stochastic approach capturing infrastructure unavailability
Weik, Norman; Nießen, Nils (Thesis advisor); Bohlin, Markus (Thesis advisor)
Aachen (2020) [Dissertation / PhD Thesis]
Page(s): 1 Online-Ressource (xx, 192 Seiten) : Illustrationen, Diagramme
Capacity planning of railway systems is a vital aspect of the rail traffic planning process focusing on the interplay of traffic and service quality. It encompasses two major, related questions: Assessing the service quality associated with given traffic and timetable concepts, and determining the traffic volume that can be operated on a given infrastructure such that attractive performance and service quality are upheld. To fulfill these tasks, a detailed understanding of both the timetable structure and the underlying infrastructure is required. In recent years, much effort has been dedicated to devising robust, passenger-friendly timetable concepts. As a consequence, capacity planning has been focusing on timetable-centered approaches such as the UIC Code 406 methodology or optimization- and (max,+)-techniques for timetable stability analysis. The role of the infrastructure itself has received less consideration, yet its reliability and its resilience in the event of failures is an essential requirement to ensure a high quality of operations. Lengthy planning procedures, a variety of different stakeholders and high investment costs require a long-term planning horizon for infrastructure design and modification. At the same time, renewal cycles easily exceed 30 years -- a period typically covering several major timetable revisions -- such that infrastructure investment decisions have long-lasting implications on the shape of the network. As a result, a certain robustness of the infrastructure with respect to uncertainty in timetable structure and service demand is highly desirable. Stochastic and queuing-based approaches, which rely on an abstracted traffic-flow view on railway operations, have been applied successfully in this strategic planning context. Individual system components such as railway line segments, stations, junctions and route nodes are interpreted as service units, for which performance parameters including waiting times, blocking probabilities or queue lengths are determined. The performance of different infrastructure designs can thus be assessed and compared as a function of the overall traffic concept, but independent of the exact structure of the timetable. To allow for general service and arrival statistics, existing approaches are typically based on mean value results. This limits the predictive quality of these models as the effects of rare events or the duration and height of non-compliance with predefined service standards cannot be considered. This is particularly relevant in the context of infrastructure failures, which have a decisive effect on punctuality and service quality, yet typically are only incorporated implicitly as a source of variation or primary delays. An explicit consideration of the state of the underlying infrastructure in the capacity modeling process is missing. In this thesis, a versatile and efficient modeling approach based on Quasi-Birth-and-Death (QBD) processes is presented, which generalizes existing queuing-based approaches for capacity analysis. By adopting a state-space representation, distributional information and new performance metrics such as the excess-probability of a predefined service quality level become accessible for analysis. In addition, it is shown how more realistic service processes including dependencies between successive train runs can be studied, such that a more detailed representation of the actually observable effects in train operations can be achieved. In making service times depend on an environment process describing the state of the underlying infrastructure, an integrated approach for railway performance and infrastructure availability modeling is developed. The entire infrastructure failure and repair process is incorporated, including the transition phases after a failure has occurred or operability has been restored. Our method hence goes beyond scenario-based approaches, where the effects of infrastructure unavailability are studied by analyzing different system settings in isolation. Using an explicit state-space representation, the criticality of individual components with respect to the dependability of service quality is assessed. Unlike in existing heuristic approaches, the effects of asset management decisions can thus be studied directly referring to a quality metric. In particular, directions for availability improvement and infrastructure modifications can be identified and assessed in terms of their respective improvement of service quality. It is shown how the methodology can be embedded in a modular, fully automated toolchain for integrated capacity and reliability analysis only requiring standard infrastructure and train data exchange formats as input. In this context, efficient model setup and solution techniques based on Kronecker-products and iterative Krylov-subspace methods, which are well-suited to cope with large sparse matrices, are made available for railway performance modeling for the first time. This allows for competitive use of the approach in rail traffic planning and asset management. Based on realistic application scenarios it is shown how the approach allows to assess infrastructure criticality on route-node scale and -- upon suitable aggregation -- for entire railway lines. The suitability of the underlying infrastructure is analyzed based on service-related performance metrics. Track modifications, asset management strategies and maintenance policies can now not only be assessed with respect to the infrastructure reliability, but also regarding their compliance with service intentions. As a result of the integrative approach to infrastructure and traffic planning, a better understanding of the interdependency of asset management and railway operations planning is achieved. This, for instance, allows to balance the reduced failure risk in a lean infrastructure with fewer tracks and switches to the additional routing flexibility and resilience of train operations in the event of failures in case of more complex infrastructure designs.