Railroads face major challenges in the coming years, smart use of data and IT can help deal with them. One way to look at this is backcasting: starting from the desired situation and seeing what tools you can use to achieve it. InTraffic developed a number of data-driven tools that can support dispatchers, among others, using Machine Learning and Artificial Intelligence.

"Data is increasingly being used as decision information. And we provide the tools to bring in data," says Marc Kooij, data solutions consultant at InTraffic. "These smart tools provide insight into data and bring out what deviates from the average. But you notice that many users still have too little grip on applications. 'I have something here, but I can't get it out.'

Train dispatchers

The various steps from raw data to decisions are to collect source data from the various systems, then turn it into information by processing, enriching and providing insight into the data. The next step is preventive management by visualizing the data and building in alerts on exceeding KPIs. This is followed by the prediction of deviations or disruptions based on the data followed by prescriptive management. This is done by automatically building preventive action into the systems.

With several examples, Kooij showed how this works. In Traffic Management (PRL), ProRail's system with which the train dispatcher works, various dashboards show actual deviations and predicted deviations. Here, different areas in the Netherlands can be compared with each other at different times. Kooij: "When the prediction deviates from the actual situation on the tracks, an alarm can be set on it." This solution is now being deployed in more places within ProRail.

"ProRail Traffic Control simulates whether future changes on the track can be implemented safely. Like the ten-minute train between Rotterdam-Schiphol and Arnhem. What will happen if we run a train on that corridor every ten minutes?

- Marc Kooij - Data solutions consultant InTraffic

Imaginary tool

InTraffic's data team likes to work from the principle of 'Backcasting'. This counterpart to predictions based on the current situation involves thinking from a desired future and the ways in which it can be achieved. "This makes it easier to arrive at system jumps," he says.

One example of a system that came about based on Backcasting is the imaginary tool Mutation Solver. With this, InTraffic aims to support dispatchers by reducing differences in workload, for example. "Especially during disruptions, a lot comes at them at once and many actions are needed at once. What you want in a disruption is that the rest of the timetable is not affected, by dampening the delay. The existing plan needs to be updated. For example, by holding up delayed trains a little longer or possibly by giving a new route around the defective equipment or infra. This system takes that work off your hands by making recommendations based on data from the track."

Renovation around the track

Another example is the smart renovation of ProRail's Process Control. Commissioned by ProRail three years ago, InTraffic began the software conversion to move PRL from the outdated OpenVMS to Linux. Since then, four of the thirteen traffic control stations have been transitioned. "We created models for this to provide insight in advance that the new software will also perform well on the Linux hardware."

"For a non-rail customer, we once looked at how to use data to predict employee vacation behavior," Kooij says. "This company noticed that planned monthly budgets were no longer being met and suspected this was due to employee vacation behavior. What we then did was to analyze the age of the employees; it turned out that in recent years they had become much younger, were more likely to have no children and thus not be dependent on school vacations. As a result, there were also far fewer on vacation during those periods, but in other periods of the year. You could apply this analysis for different purposes."

Load on systems

Back on the tracks, there are more telling examples of achieving a desired situation through data and the proper use of it. "At a large traffic control station, systems were overloaded since the corona outbreak because controllers were no longer sitting in pairs behind the same workstation, but each separately. By running an analysis on the CPU load, the load on the systems, we found a way to better distribute this load among the different components."

We are also working as a data team on the success of the Collate program. "ProRail Traffic Control simulates whether future changes on the track can be implemented safely. Like the ten-minute train between Rotterdam-Schiphol and Arnhem. What happens if we run a train on that corridor every ten minutes? And if we simulate a breakdown? How does it affect the rest of the timetable, or if we put in an extra train? We have experienced dispatchers work with the simulator to see what the effect is, how they deal with it and what measures still need to be taken before a change goes live. After such a simulation with domain experts, we make the punctuality and customer inconvenience visible on dashboards in near real-time."