High-Speed AI: Hitachi Rail Advances Real-Time Railway Analysis Using NVIDIA Technology

NVIDIA announced that Hitachi Rail is integrating NVIDIA AI technology to lower maintenance costs for rail operators, reduce train idling time and improve transit reliability for passengers.

The company is adopting NVIDIA IGX — an industrial-grade, enterprise-level platform that delivers high-bandwidth sensor processing, powerful AI compute, functional safety capabilities and enterprise security — into its new HMAX platform to process sensor and camera data in real time.

By removing the lag time between data collection and analysis, the HMAX platform will enable Hitachi Rail clients to more quickly detect train tracks that need repair, monitor the degradation of overhead power lines and assess the health of trains and signaling equipment.

Hitachi Rail estimates that proactive maintenance costs around 7x less than emergency repairs done after infrastructure fails unexpectedly. Its existing AI monitoring systems are already reducing service delays by up to 20% and train maintenance costs by up to 15% — and are cutting down energy consumption by decreasing fuel costs at train depots by up to 40%.

With real-time analysis using NVIDIA IGX and NVIDIA Holoscan platform for sensor processing, the company aims to further increase these savings.

“Using previous digital monitoring systems, it would take a few days to process the data and discover issues that need attention,” said Koji Agatsuma, executive director and chief technology officer of rail vehicles at Hitachi Rail. “If we can instead conduct real-time prediction using NVIDIA technology, that enables us to avoid service disruptions and significantly improve safety, reliability and operating costs.”

Building on its existing collection of HMAX applications — which are currently running on data from 8,000 train cars on 2,000 trains — Hitachi Rail has used NVIDIA IGX and the NVIDIA AI Enterprise software platform to create new accelerated AI applications to help operators monitor train fleets and infrastructure. NVIDIA AI Enterprise offers tools, pretrained models and application frameworks to streamline the development and deployment of production-grade AI applications.

These applications, available soon through the HMAX platform, can be used by the company’s international customer base to process huge quantities of data streaming from sensors onboard trains, taken from existing systems or imported from third-party software already in use by the customer.

In the U.K., for example, each Hitachi Rail train has sensors that report nearly 50,000 data points as frequently as every fifth of a second. AI infrastructure that keeps pace with this data flow can send train operators timely alerts when a component of a train or rail line needs maintenance. The AI insights can also be accessed through a chatbot interface, helping operators easily identify trends and opportunities to optimize maintenance schedules and more.

“If a potential issue isn’t identified and fixed promptly, it can result in a service disruption that causes significant economic loss for our customers and impacts the passengers who rely on these transit lines,” Agatsuma said. “NVIDIA AI infrastructure has enabled us to get immediate alerts on thousands of miles of railway for the first time, which we anticipate will reduce delays and disruptions to passenger travel.”

The opportunities go beyond monitoring trains and tracks.

By mounting cameras atop trains, Hitachi Rail can monitor power lines overhead to identify degrading electric cables and help prevent disruptive failures. Traditionally, it takes up to 10 days to process one day’s worth of video data collected by the train. With NVIDIA-accelerated sensor processing, data can be processed in real time at the edge, sending only relevant information back to operational control centers for analysis and action.

Source: NVIDIA media announcement