| Popis: |
Summary We present in this study results of the application of Machine Learning techniques to Distributed Acoustic Sensing (DAS) records. Our work is focussed on tests lead in the FEBUS OPTICS’s test centre located in Pau, South-West of France. This facility is equipped with a 22m-long buried pipeline instrumented with different fibre-optic cables: Single-Mode (SM), Multi-Mode (MM) fibres in tight or loose tube configurations, fibre cables directly buried in the ground or inside conduit and located at various distances from the pipe. Different events were simulated along the fibre (footsteps, compactor, vehicle, etc.) and simultaneously recorded with the different kind of fibres. The classification is run, using the Random Forest supervised algorithm, in order to identify the different events. We obtain a good quality of the classification with an accuracy equal to 91,82% emphasizing the good efficiency and reliability of our Machine Learning algorithm. |
