InnoTrans: Shifting away from using dedicated test vehicles to monitor railway infrastructure in favour of attaching sensors to revenue-earning trains could increase the amount of data collected and processed ‘one-thousand fold’.
This is the view of Koji Agatsuma, Hitachi Rail’s Chief Technology Officer for Vehicles. Speaking to Railway Gazette International at InnoTrans on September 25, Agatsuma said the changing demographics of the rail sector were already lending fresh momentum to the transition to condition-based maintenance and asset management models.
‘We used to make younger staff fill out reports and compile spreadsheets of data at the depot — this approach has to change’, Agatsuma explained. ‘By using digital tools and a more “gamified” environment, the work becomes more attractive to new entrants.’
HMAX launch
To support the use of revenue-earning trains to collect condition data for rapid analysis underpins, the company launched a new product line at InnoTrans. Hyper Mobility Asset Expert is intended to cover applications across rolling stock, infrastructure and signalling. Hitachi Rail explained that HMAX would create a digital twin of the overall rail ecosystem, incorporating sensor data and feeds from core components, additional sensors and third parties.
On September 26, Hitachi Rail announced that København operator Metroselskabet would be the first company to deploy HMAX in a customised format. The technology is to be rolled out across lines M3 and M4 of the driverless network under a multi-year agreement.
The contract will see Hitachi Rail install sensors on the metro trains, including a vibration monitoring system, and integrate the real-time data into a bespoke version of HMAX. The sensors will supply data on the performance of the trains’ subsystems, bogie and wheelset health, as well as the condition of the track.
Faster processing is key
Many rail and metro organisations have had aspirations for years to base maintenance interventions on condition rather than time. But Agatsuma explained that a lack of high speed data processing capability had held back progress.
‘Many railways like the condition data they receive, but the most common problem is that it is not delivered fast enough’, he said. ‘I often hear that “we wanted our notification of a problem yesterday, not two weeks from now. Why were we not told earlier?”.’
Faster chips for data processing and advances in edge computing have led Agatsuma to believe these issues could now be resolved. On its stand, the company had a live demonstration of a roof-mounted camera monitoring the overhead line equipment on a British main line. Within 15 sec of the visuals being recorded by the camera, an AI-augmented analysis of key parameters was displayed on the adjacent screen.
Hitachi Rail said that prior to the deployment of the AI enhancements, it could take up to 10 days to process all the data that a train collected in a single day. According to Agatsuma, the key to rapid analytical work is edge computing, underpinned by high speed processing power. Hitachi Rail is using Nvidia’s IGX industrial-grade analytics software and its Holoscan sensor processing suite as part of a strategic partnership between the two companies. Edge computing enables data processing in the immediate vicinity of the collection point — in this case, onboard the train. Only the most relevant data is then transmitted back to a depot or control centre.
In March, Hitachi’s IoT and industrial enterprise IT division Hitachi Vantara signed a formal partnership with Nvidia under which the two firms would work to develop a new range of AI-driven software products for use across various industries.
