Technology transformation
The enabling technologies for rail transformation
In the previous pieces, we discussed the digital transformation of rail, from engineering design and manufacturing, to digitizing trains and track, to optimizing maintenance and operations and creating new digital customer services.
Whilst we have looked at these from the point of view of innovation across various areas of the rail industry, these areas require overlapping technologies and IT infrastructure, and will benefit from different rail divisions working together. Sensors will help automate train schedules, but also provide valuable data to help engineers design a new generation of trains built around the reality on the track.
Designing these solutions will rely on a range of similar digital technologies and approaches, which must often be joined up and designed to talk to each other. This final article explores the enabling technologies that will deliver rail transformation, and how to approach them.
Digitize your world - Sensors
In order to digitize, trains are going to need to deploy a wide range of sensors. Imaging such as LiDAR and cameras will collect the data that allows the digitization of the train and the track. Movement sensors, like GPS and accelerometers, will provide real time information about the train and allow real time responses to emergencies and generate data from which to build AI models that optimize driving for safety and efficiency. Vibration, temperature, and door sensors will detect early signs of problems that can be flagged to predictive maintenance systems.
That is to name but a few of the myriad sensors that future digital trains will deploy.
Deploying sensors at scale will come with costs and skills implications. Rail companies need to decide on the right sensors for their fleet – which will contain trains of varying ages and uses – looking at cost/benefit before rolling out.
But they shouldn’t rush into connecting everything for the sake of it. There are workarounds if you know what you are doing. For example, data from cheap vibration sensors can be used to infer a great deal of detail about train and track conditions, without the need for pricier systems like LiDAR on every train.
Most importantly, deployment should be done with scale in mind, ensuring all sensors work to the same standards so data can be combined, and future upgrades can be added or swapped out without needing to upgrade the system.
Bring everything together - Cloud storage
All of this data will need to be processed and analyzed to generate insights. Some will be fed straight into real-time systems like braking or driver assistance. Some will be directed into predictive maintenance systems back at the depot. Some will sit in databases for analysis by engineering and data teams looking for new opportunities for optimization.
This needs a place to store data, and the capabilities to process it, draw insight from it, and deliver those insights to the people and automated systems that need them. The most obvious way to do that is a hosted cloud environment, such as those offered by Tech companies – a single space where everything can be stored and managed, and where there are readily accessible analysis tools to build models. These can then be connected to onward systems – such as engineering design tools, maintenance scheduling, or enterprise resource planning software.
An alternative could be to build such an environment on your own servers, which is harder to do, but brings greater certainty over where your data is stored, which may be valuable for sensitive data.
These environments need to be carefully set up with standard approaches to data management, to ensure data can be easily found, and different data feeds can be easily combined - a ‘Single Source of Truth’ in IT parlance. Ultimately, as data collection and model building increases, this will become a digital twin of the train and the rail network.
Real time responses – Embedded software
Some onboard devices – such as emergency brakes - will need to react instantly. They cannot afford even the brief delay of sending data back to the cloud, or the risk of a momentary loss of connectivity. These devices will need their own onboard processor and embedded software – programs designed to take data, run it through a model, and deliver a resulting instruction on the device.
These will need to be carefully designed and tested in their own right according to the high standards of safety critical industries. But again, they should be designed with the whole ecosystem in mind, since away from emergencies, such data can be immensely valuable to other applications – from predictive maintenance to engineers designing new braking systems - and so needs to be able to easily flow between joined up systems.
Smart decisions, accurate insights: Analytics, AI and Gen AI
If you do the above right, you will find yourselves with rich repositories of data, from which you can draw insights and design intelligent software, using analytics and AI. These tools allow you to analyze vast amounts of sensor data for complex patterns, beyond human comprehension, to do things like predict maintenance needs, and optimize train schedules and routes.
Generative AI offers a whole new suite of opportunities. It can be used to absorb vast amounts of uncurated information – such as video and instruction manuals – that would not traditionally be thought of as data, and understand its context. This allows teams to develop tools that can quickly propose a set of useful answers – how to fix a machine for example – or even intuit the right answer from different pieces of information, or apply one solution to a different problem, even if the question is not specifically addressed in its training material. Of course these tools need guardrails and governance structures to ensure they are used appropriately in a context that demands precision and rigor. We recommend a Hybrid AI approach [13].
Not everything needs to be hyper intelligent, but now things can be. Doing this well is about having the right people who can choose the right (and most cost-efficient) tool for the job.
A report [8] recently produced by the International Union of Railways (UIC) in partnership with McKinsey concluded that increased AI adoption could add roughly $13 to $22 billion of value a year to the sector. That report found that about 20 AI use cases were currently being explored, and concludes that AI could present an opportunity of around €700 million a year for a €5 billion rail company.
On a related note, the European Infuture Institute, said the following in a recent article in International Railway Journal: [9] “AI is already transforming the rail industry through the automation and optimisation of every aspect of operations, including traffic control and maintenance. It is also a great tool to enhance the efficiency and reliability of rail transport since it strongly supports the development of intelligent rail infrastructure management systems.”
Think global and local
New trains will have to better serve the global environment. But these trains, for infrastructure reasons, must also live with existing local rules and technology - even in cases where some regions (eg. the EU with its ERTMS signaling rules) are trying to standardize aspects of rail operations. In many cases, today’s trains are too country specific.
Instead, these trains must be designed as a more generic, modular product that can be easily modified to meet the diverse requirements of different regions, operators and roles. As such, flexibility in the design and the ability to adapt to change is key to success.
Connecting the train – 5G
Finally, the onboard devices need to connect directly and continuously to the cloud. The onboard wifi needs to have a reliable, high bandwidth connection to enable the kind of services customers increasingly expect, such as high-definition video calls, and in future, perhaps VR.
Right now, most trains are not able to deliver such connectivity. The solution will be the Future Railway Mobile Communication System (FRMCS), the upcoming global standard designed to replace the existing 2G-based GSM-R (Global System for Mobile Communications - Railway) system. FRMCS is based on 5G technology and utilizes 5G’s high data rates, low latency, and increased connection density to deliver enhanced voice and data communications on trains – for people and IoT devices – and seamless data exchange between trains and railway infrastructure. It is highly scalable, setting trains up for ever higher data demands from passengers and digital trains.
Technological visions are all very well. But the reality is that these are technically complicated, especially when retrofitting to a safety-critical industry, with a wide variety of moving and fixed assets, all around the country (and often crossing borders), of varying ages, most of which were not built with all this digitization in mind.
It is therefore important to tread carefully. Work with companies that have ‘been there, done that’ and learned the lessons along the way. Take advantage of battle-tested methodologies for rail technology selection, deployment, and systems integration. Utilize frameworks for software development to ensure both speed and consistency in the
development of the myriad different software systems that will be needed, both embedded on devices and in the cloud.
In doing so, engage with all country regulatory and homologation bodies. The future of rail needs to be consistent and trans-national. When a new train rolls onto the production line, we want to know all its systems will work anywhere in the world, and any future technologies can be easily implemented into its IT architecture so that they seamlessly work with everything else on the network. That needs standardized approaches – or at least reliable ways of connecting different systems – not just across the company, but across all companies, in all countries.
All of this needs to be wrapped in excellent project management. Keep your eyes on the goal, and break down progress into incremental steps with small wins along the way. Rail is safety-critical infrastructure and rightly conservative, so we need to go slowly, and prove concepts rigorously with attention to safety. But we also need to show that – through digital technologies – there are better answers out there that make rail safer, more efficient, more sustainable and more attractive to customers.
