Background

The weaknesses of rail freight traffic —such as long transportation times, low flexibility, and limited forecasting accuracy—prevent it from reaching its full potential. This is particularly evident in the European Single Wagonload System (SWLS), which is designed to meet the shipping needs of customers who require flexibility in shipment volumes and schedules. As a result, the transport capacities of trains are not used efficiently. The introduction of a modern, digital Single Wagonload (SWL) system could address these inefficiencies and would support the goals outlined in Austria’s 2030 Mobility Master Plan and the Masterplan Güterverkehr 2030, which aim to increase the modal share of rail freight to 40%.

Project Content

The Austrian Single Wagonload System (ASWLS), like many in Europe, operates as a hierarchical hub-and-spoke network, where wagons are picked up at industrial sidings, routed through local hubs and shunting yards, and delivered to customers after switching train connections several times. That SWL providers do not operate the entire system poses a major challenge. It leads to fragmented management and one consequence are inaccurate Estimated Times of Arrival (ETAs). To address this, a holistic approach based on an operational Digital Twin of the ASWLS is proposed, that combines RCA-managed system models with data-driven models of external system components.

Goals and Main Steps

We intend to design a model-based digital representation of the Austrian Single Wagonload System (ASWLS). This encompasses two major steps:

Simulation and Scenario Assessment

Operational Digital Twin

We set up of a predictive model that leverages data from the digital twin, along with other sources like real-time data and forecasted demand. The goal is to improve the prediction of Estimated Time of Arrival (ETA) for wagons at final destinations and key points of interest (POIs), such as shunting yards.

We test the model’s capacity to support planning and management decisions. For instance, assessing the effects of timetable rescheduling on individual transport orders.

Methods

The accuracy of Estimated Time of Arrival is affected by delays, differing planning approaches between Rail Cargo Austria and shunting yard operators, and inefficient communication among the parties involved. To address this, we are developing a Digital Railway Freight Model (DRFM) that integrates operational data—such as yard arrival and departure times, train movements, and booking data—with predicted near-future demand from a data-driven model. The DRFM combines a macroscopic rail network model with machine learning-based surrogate models to enhance ETA predictions and support improved operational and tactical planning decisions. Additionally, a machine learning-driven shunting model aids in assigning wagons to outbound trains and can deal with dynamic changes and varying shunting yard operations across Austria when predicting departure times.

Results

META TRAIN pursues a novel approach that paves the way for more accurate ETA predictions. By using modern methods like machine learning and digital twins, we enable flexible and dynamic carriage-to-train assignments and streamline shunting processes. The project supports the digitalization and modernization of rail transport, making it a more attractive alternative to road transportation, while increasing rail freight's modal share and contributing to sustainability goals as outlined in Austria’s 2030 Mobility Master Plan and similar initiatives.