Machine Learning Techniques Deliver Railway Application Improvements

Commuter train in station

OptaSense’s latest generation QuantX Distributed Acoustic Sensing (DAS) interrogator unit delivers long-range quantitative data performance with high-fidelity and sensitivity – enabling the potential for numerous advanced railway applications to be delivered. With an operational range up to 50 km, the IU is capable of transforming long fiber optic cable lengths into a dense sensor array – producing in real-time, a spatially and temporally rich dataset required for many railway applications.

However, with this increased dataset, challenges are presented in the form of high-volume data handling, processing and classifying signals of interest (such as rolling stock, track features or intrusions) efficiently.

To harness this power while addressing the challenges posed by the increase in data volumes, OptaSense is utilizing a number of Machine Learning (ML) techniques to maximize the value from the improved data fidelity and reduce QuantX commissioning times. A key area of focus for our Transportation team has been using supervised ML techniques to deliver a step-change improvement in the detection of rolling stock movements which enables a performance improvement across the rail solution portfolio. In addition to supervised ML techniques, another focus area is to employ adaptive classification algorithms which utilize goal seeking methods to dynamically and continuously tune every channel along the fiber according to assessments of background noise levels – all day, every day.

Read the full paper here.

Scroll to Top