Objective
The customer, a company providing traction rolling stock maintenance, requested the creation of a tool for operational management of train and locomotive movement on a railway network section, enabling the verification and generation of proposals for locomotive traffic management.
The customer's goals were:
- minimizing the risks of locomotive non-delivery to a train;
- determining the optimal number and geographical distribution of locomotives to balance the risks of train non-removal and fixed locomotive costs;
- identifying the factors behind deviations from optimal parameters in order to develop corrective measures.
Data Analysis
To build the model, we analyzed the customer's data on operations within the railway network section for November 2018: statistics on trains removed and not removed, data on locomotive movement, the number of repair depots and their capacity, regulatory documentation on repairs, and data on repairs actually performed. The structure and capacity of the railway network itself were also analyzed.
Based on the data analysis, 4 locomotive depots, 5 locomotive types, 5 categories of repair work, and the specific features of the railway network section under consideration were identified: weight zones, pushing sections, and single-track sections. Repairs were divided into scheduled and unscheduled. The average frequency and duration of scheduled repairs were estimated based on regulatory data, while the frequency and periodicity of unscheduled repairs were estimated from the customer's actual data.
The model needed to reflect:
- the actual mileage and condition of the locomotive fleet at the start of modeling;
- the division of the railway network section into depot service zones (locomotives are re-coupled at zone boundaries);
- the dependence of the number of locomotives assigned to a train on the train weight;
- the ability of a locomotive to operate in different modes:
- at the head of the train;
- multiple-unit traction;
- pushing;
- the weight zones of the sections – when crossing a weight zone boundary, a new locomotive is requested if necessary.
Model
As a result of the data analysis, we created a simulation model of traction rolling stock movement on a railway network section.
The logic of locomotive crew operations – shifts, locomotive re-coupling, idle time – was implemented using agent-based and discrete-event modeling. Using agent-based modeling, we reflected the domain entities in the model – trains, locomotives, and depots. The dispatching and locomotive assignment process was reflected using flow chart and request tools. To define the railway network, we used the JGrapht library – the network section was defined using an undirected graph, and Dijkstra's shortest path algorithm was used to assign a route to a locomotive.
As a result, the model features a convenient graphical interface for displaying locomotives on a GIS map, their statuses, and a reporting generator in a format similar to the business customer's report format.
The model allows calculating the availability factor (KIO) and technical readiness factor (KTG) of the locomotive fleet by depot and locomotive series, and provides the ability to load locomotive movement scenarios in a specified format.
The graphical interface, output report, the ability to load desired locomotive movement scenarios into the model, and the built-in calculation of the locomotive fleet's KIO and KTG – all of this makes the model a convenient tool for evaluating proposed corrective initiatives and for operational management of the locomotive fleet.
Result
As a result of the project, the customer received a tool that enables real-time adjustment of locomotive crew operations to reduce locomotive idle time and delays in locomotive delivery to trains.
You can explore a simplified version of the model in the public AnyLogic Cloud.