Optical Fiber Networks Can Keep Rail Networks Safe

This article is part of our exclusive IEEE Journal Watch series in partnership with IEEE Xplore.

Rail networks are vast, which makes it difficult to conduct comprehensive, continuous safety monitoring. Researchers in China have suggested analyzing the vibrations of existing fiber cables buried underground alongside railway tracks to detect problems.

In a study published 5 March in the Journal of Optical Communications and Networking, the research group demonstrated through experiments how the technique can successfully identify a number of issues associated with train safety, including faulty train wheels and broken sound barriers alongside the railway tracks.

Sasha Dong is a junior chair professor in Southeast University’s School of Transportation, in Nanjing, China. She notes that traditional approaches for monitoring railways—such as video surveillance, radar, and ultrasonic sensing—can be effective, but they are often limited to monitoring railways at single points along entire systems.

“As a result, they are not well suited for continuous coverage along an entire railway line and are also more vulnerable to weather conditions, environmental factors, and power supply constraints,” she says.

Instead, Dong, Yixin Zhang at Nanjiang University, and their colleagues used a technique called distributed acoustic sensing (DAS) to analyze the vibrations of underground optic fiber cable alongside railway tracks to detect safety issues. Specifically, pulsed light is sent along the cable, and the propagation of scattered light is used to detect and quantify vibrations along the cable.

The researchers developed AI models to filter out the noise from those signals and to identify the particular vibrations associated with various kinds of unsafe conditions, such as damaged or defective wheels.

Dong notes that railways already have extensive optical fiber networks for communication buried underground alongside them, meaning that the cables can be harnessed as a sensing medium with no extra power supply or need for another expensive network to be constructed. Instead, monitoring stations could be installed at intervals along the railway track, with extension cables connecting a DAS system to the main cable.

To develop their DAS system, the researchers set about collecting data on different railway safety issues and training machine learning algorithms to identify specific vibrations associated with each one.

For example, they trained a model to detect the trajectory of trains using DAS data. This involved more than 13,000 samples of trains moving along tracks, where their direction was confirmed using data. This model achieved an accuracy of 98.75 percent.

In another endeavor, the researchers took samples of a train with wheel-pair faults—where there is damage or a defect on the railway wheels or their connecting axle—moving along a 60-kilometer stretch of railway track in Kunming, Yunnan, China. The researchers were able to clearly detect when there was an issue: The vibration frequencies of normal wheels were mainly concentrated below 60 hertz, while the frequency of faulty wheels could get as high as 100 Hz.

DAS may also be useful for detecting problems with sound barriers, which are the paneled walls on either side of the railway track that reduce the sound of trains as they pass surrounding neighborhoods. The researchers removed the rubber paneling from sound barriers to simulate faulty barriers and repeatedly struck the barrier with a rubber hammer, using the resulting sound data to train another model. This model could accurately detect faulty sound barriers with 99.6 percent accuracy.

The team also explored how well machine learning algorithms could detect abnormal events along the railways, such as humans climbing over trackside fences, rocks falling on the track, illegal construction activity such as excavator operations, or other environmental disturbances. These types of events were a bit more difficult to distinguish at first, but by feeding a lot of data into the model, the researchers were able to boost the model’s accuracy for these types of events to 97.03 percent.

These results suggest that DAS has the potential to be an effective tool for monitoring railway systems. “What we have found most surprising is that a single, existing fiber deployed along the railway, with appropriate modeling and algorithm design, can support so many different monitoring tasks at the same time,” says Dong. “This kind of multipurpose use of one fiber system has strong engineering value.”

Dong acknowledges that these experiments were done in controlled environments and emphasizes the need to capture more vibration data under real high-speed train operating conditions. Nevertheless, she says, “the results of this study suggest that this [approach] is feasible and has strong potential for practical application.”