The oldest rail network in the world, with over 32,000 km. of track, is investing in artificial intelligence and machine learning to streamline maintenance and deal with weather-related challenges.
Two months ago, during the AI and Big Data Expo, I had the opportunity to meet Nikolaos (Nick) Kotsis, Chief Data Scientist at Network Rail. At the conference, he was one of the speakers, talking about their digital transformation and how Network Rail was implementing new ways to inspect and manage the network’s assets, shifting work from traditional planning and maintenance schedules to a proactive “predict and prevent” approach.
At the time, Kostis mentioned the data collection happening in real-time, simultaneously from the trains using the network, the people inspecting the tracks, drones, helicopters, and over 30,000 IoT sensors deployed all over the country.
All this data collection allows Network Rail to know what is happening and take immediate action when something goes wrong or needs fixing. But the real magic, which helps predict and prevent incidents, and provides predictive maintenance, happens when AI and Machine Learning are applied to that massive amount of data.
To learn more about how Network Rail’s Data Science department works and how it impacts the organization, we reached out to Nick Kotsis again. He answered our questions by email.
PV: As discussed in our previous conversation, Network Rail is undergoing a substantial digital transformation in the field and the data center. Can you tell us a bit about your data science department and its role within the organization?
Nick Kotsis: Our initial plan for the data science function was leaning towards being primarily guidance and advisory; however, once we began engaging with customers, we realised that taking on responsibilities for delivery would not only return cost benefits to the taxpayer but would also help our partner network operate more effectively.
Inspiring confidence and gaining the trust of our customers and suppliers was the focus of my role in the first six months. Since then, we have evolved to become an actual delivery function for data analytics, advanced machine learning, and AI technology packaged into fully integrated digital products that customers can use with minimal training. The end result is a trusted service supported by a skilled, confident team focused on the customer and responding to the most complex problems in Network Rail.
When customers from across the organisation ask for help, it means we do something right.
PV: You said before that NR is a complex machine, managing the network and some of the largest train stations and maintaining the freight trains. How is data analytics (and AI) affecting the different services?
Nick Kotsis: Our organisation is responsible for maintaining a complex infrastructure that is vulnerable to environmental and weather conditions and the constant pressure of hundreds of daily train journeys.
Network Rail maintains 20,000 miles of track. Our job is to maximise the use of data to make the maintenance of the infrastructure a safer and more efficient environment for both our passengers and workforce.
To give you an example, performing remote inspections on track assets with the assistance of AI technology instead of visiting the track is a genuine safety benefit for our maintenance teams. Of course, we are not planning to pause physical inspections, but if we could confidently limit them to absolutely necessary ones, that would be a true benefit.
Enabling remote inspection using AI technology
Data analytics and the more sophisticated machine learning techniques consistently demonstrate high quantitative and qualitative benefits. Examples of these benefits can be seen in: the prediction of incidents; the automation of tasks that would be repetitive and mundane for a human; the complex risk assessment on thousands of assets in a split second; and complexity management using optimisation algorithms which also bring speed to complex decision making.
In our case, we have successfully developed automated risk assessments using computer vision techniques that identify assets for immediate attention on the track and surrounding areas. We also established a preventive maintenance process driven by predictive algorithms which calculate the likelihood of an asset failing days or weeks before it actually happens, allowing us to resolve an issue before it becomes a problem. Both of these systems offer significant improvements to safety along with reduced delays and disruption for passengers.
The advanced big data engineering and algorithmic (AI) logic we use behind the scenes should, and will, continue to expand across every part of our infrastructure. I am confident that eventually, we will reach the desired levels of deployments needed to scale up our operations to prevent the incidents.
PV: The UK has the oldest rail service worldwide. Nevertheless, across Europe and many other places are also “old” rail services operating, each with complex and specific challenges. Based on your experience, what would you recommend to those organizations starting or undergoing digital transformation?
Nick Kotsis: Finding the correct answer to this is not easy as every organisation will be starting at a different point and has a different maturity trajectory. Also, financial investment for some organisations can be made easier than in others, and thus the digital journey will be very different.
As with most complex initiatives, digital, data, and AI success depends on leadership and commitment to completing the journey. In our case, we are fortunate to have solid technical leadership from our CIO and CTO. They saw our destination at the early stages of development and supported us in making the data science vision a reality.
My advice to colleagues in other organisations? be clear about the vision and data strategy, engage with customers who need your help, create the right team and set them for delivery, and focus on projects with clear benefits.