Litherland, Jack J.M.
(2019)
Whole system approaches to railway asset management.
PhD thesis, University of Nottingham.
Abstract
As the world population continues to grow, the demand for rail transport increases. The requirements on railway networks already at capacity continues to expand. In 2017-2018 there were 1.7 billion passenger journeys made on the UK railway network, compared to just over 700 million in 1994. Yet the number of track kilometres in the UK has remained near constant over that same time period. These changes are making maintenance scheduling more challenging; increased usage is leading to faster degradation yet, with more trains per hour and services starting earlier and finishing later, the time available to perform maintenance is shrinking. In many areas, it is extremely difficult to build new railways as land is both limited and expensive, additionally there are a range of legal challenges. Therefore innovative approaches to railway asset management are needed to maximise the performance of the existing network.
This research will investigate how whole system approaches to railway asset management, which consider whole system condition and whole life cost, can be used to make maintenance as streamlined and efficient as possible and tackle these challenges. A range of models have been created to aid railway asset management including Markov models and Petri net models. Nonetheless, previous models have generally only considered a single asset class and only considered a small section of the railway. The aim of this thesis is to investigate the feasibility of a framework, which allows a range of different railway assets to be modelled simultaneously, over a large section of railway.
In order to create a whole system model this thesis begins by developing Petri net models for a range of assets within the railway network namely: track (rails and ballast), sleepers, switches and crossings (S&C), tunnels and the signalling system. The asset models were developed based on Network Rail data, literature and expert judgement.
The Petri net models for the various assets were then converted into modules and combined using the hierarchical Petri net (HPN) framework, to form a system model. The hierarchy proposed contained three levels, allowing different phenomena to impact the system in different ways. The system model enabled the effects of dependencies between assets to be explored as well as allowing opportunistic maintenance to be considered. The framework was constructed in such a way as to allow additional asset classes to be included, in the framework, in the future.
The system model was applied to the Bletchley Delivery Unit, a busy railway corridor, as a test case. The test case contained; a master/control module, 849 track modules, 849 sleeper modules (705 concrete, 53 timber, 91 steel), 61 S&C modules, 21 tunnel modules and a signalling module.
Due to the size of the whole system model, the simulation time and memory required needs careful consideration. The final part of this research explores a range of means to speed up the simulation of Petri nets, and reduce the memory required to solve them. A range of techniques for streamlining the Petri net algorithm were explored. The research also investigated the benefits of using parallel processing to solve Petri net models. Heterogeneous programming, which combines the central processing unit (CPU) and the graphics processing unit (GPU) was also explored. It was concluded that the memory requirement is likely to limit performance before computation time, and also limits the benefit that can be achieved from heterogeneous programming.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
![[img]](/style/images/fileicons/application_pdf.png)