Rail, as an industry, has evolved at a fairly consistent rate since the industrial revolution. Now this gentle advance has accelerated dramatically. Advances in computer science and data analytics are transforming the way we keep railway systems fully available and operational, delivering dramatic improvements to safety, performance, and cost-effectiveness. Computing devices embedded in trains, signalling systems, and railway infrastructure can now send and receive data online. Alstom is in the vanguard of this ‘Internet of Rail’.
In order to avoid the unavailability of trains or other elements of signalling or infrastructure, rail companies have traditionally adopted a preventive approach towards maintenance. In other words, equipment is serviced regularly to lessen the likelihood of it failing without warning. But advanced analytics offers an alternative, enabling us to foresee with precision when such parts would reach the end of their useful lifespans – and schedule servicing works accordingly.
Alstom started its condition-based maintenance programme in 1998. Eight years later, the company started to transfer data from the train to the ground. Today, Alstom is leading the evolution in rail services with tools such as HealthHub – its predictive maintenance solution, TrainTracer for monitoring train health in real time, TrainScanner – its automatic diagnostics portal, and TrackTracer which monitors the condition of tracks and catenaries.
Such an approach saves time and money because parts and components are only serviced when necessary. For instance, conventional wisdom says that heating system filters must be changed every three months in order to avoid risk of a breakdown. By switching from a mileage-based approach to a condition-based one, the filter’s remaining lifespan from the due replacement date can be extended by up to 90% every five-and-a-half months – all without taking any undue risks.
This technology is also vital for monitoring infrastructure. When a point machine fails, fixing the problem can be like repairing a faulty bulb in a string of fairy lights. A stretch of track 5,000km in length might contain 100 such devices. In the past, technicians would have had to examine each one individually. This costly, disruptive process, which also entailed taking trains out of service, has now been circumvented. We can pinpoint exactly which one has malfunctioned, when and why.
Full steam ahead
Three factors have facilitated this shift. First, sensors are becoming smaller and cheaper, so can be used in locations that would previously have been impossible. On trains, they can monitor all manner of functions, from filters and door locks to heating and ventilation or traction components. Not to mention signalling equipment and elements of infrastructure.
Secondly, connectivity has improved. The data the sensors gathered can be transferred online at ever-increasing speeds. Finally, computers are far more powerful, so vast data sets can be analysed to reveal patterns, trends, and associations.
Alstom places an emphasis on ‘smart data’ rather than ‘big data’, prioritising quality over quantity. For instance, while it could, in theory, place sensors all over a train, the amount of information yielded would be costly and time-consuming to interpret. Instead, the company uses its decades of engineering expertise to identify areas most in need of monitoring. Typically equipment that is particularly susceptible to wear and tear fails more frequently or causes great disruption when it goes wrong.
The quality of the raw data is critical, which is why, Alstom developed its in-house HealthHub Platform. This innovative portal pools findings relating to individual components, from brakes and bogies to toilets. A diagnostic modelling feature named Health Indicator generates diagnoses and prognoses for each machine, predicting when, where, and how future degradation is most likely to occur.
There are certain dynamics that Alstom does not model, and there will always be factors that are beyond its control. But when data experts and engineers work hand in hand, variations caused by context, such as the harsh Middle Eastern climate, can be taken into account.
At its most basic level, predictive maintenance means anticipating problems. For many years, Alstom has studied the ways in which train components deteriorate. We know the impact of each component on cost, reliability and availability, and plan accordingly. Some problems can be remedied quickly; others may be dangerous or difficult to solve. If a particular piece of equipment is very hard to reach, we can arrange the extra staffing needed to service it. If it has to be sourced from a third party, we can replenish stocks well in advance, avoiding the disruption caused by having to take trains out of service.
This approach centres on system behaviour and it is this that delivers reliability. By coupling our understanding of the physical and digital realms, we can boost savings, safety and reliability, realising the Internet of Rail’s potential.