Railway Infrastructure / Fibre-optic Sensing
Distributed Acoustic Sensing
CLIENT — Railway infrastructure monitoring company
Consultancy and development for a DAS-based sensor package that interprets acoustic signals from fibre-optic cables to monitor railway infrastructure across many kilometres of track.
The challenge
Every microphone is noisy and has its own unique response signature, making interpretation of large, complex datasets difficult. The client needed actionable insights from distributed acoustic sensing data for real-time infrastructure monitoring.
Our approach
Custom C++ for raw-data analysis (signal detection and classification using convolutional neural networks); a Qt/C++ UI for sample viewing and annotation; Docker-containerised C++ programs for DAS data manipulation, including integer-arithmetic implementations for potential FPGA translation; plus ongoing consultancy, reporting and documentation.
Results
Our analysis and visualisation tools let the client explore the data in new ways, revealing previously unseen structure in the signal — distinguishing cars driving near the fibre, trees swaying in the wind, and people walking close to the track.
Technologies
- C++
- Convolutional neural networks
- Qt
- Docker
- Signal processing
Turning noisy fibre-optic acoustic data into real-time railway infrastructure monitoring.