Every fibre-optic cable can tell a story….
Distributed acoustic sensing (DAS) works by sending pulses of light down a fibre-optic cable and detecting the returned signals scattered off imperfections along the cable length; interpreting the signals allows the strain in the cable at a point to be calculated. This allows a cable, in effect, to be used as a series of acoustic sensors, many kilometers long, like a series of microphones, albeit where every microphone is noisy and has its own unique response signature. For a cable laid alongside a railway track, there is tremendous potential for real-time observation and prediction of the state of the track, rolling stock and other assets and events, but interpretation of these large noisy data sets is a challenge.
React AI were contracted by a railway infrastructure monitoring company to provide ongoing consultancy and development work over a series of months, to help implement a world-leading DAS-based sensor package.
React AI undertook a wide-ranging collaborative role with our client, including:
- Raw data analysis via high-performance custom code
- Analysis included signal detection; classification using convolutional neural nets; and asset management/predictive maintenance.
- Creation of a custom UI written in Qt/C++, for both viewing and marking-up DAS samples
- Bespoke development: a suite of C++ programs for manipulating and analysing DAS data, delivered as Docker containers, and where appropriate implemented in integer arithmetic ready for translation to an FPGA
- General advice and consultancy; reports on work undertaken, documentation of all results, conclusions and recommendations for future stages.