2025_programme: Deep Learning Approach to Remove Distributed Acoustic Sensing Instrumental Noise

• Day: June 20, Friday
    Location / Time: A. TERPSIHORI at 09:30 - 09:50
• Last minutes changes: -
• Session: 05. Distributed Fiber-Optic Sensing technology for underwater acoustical monitoring
  Organiser(s): Alexander Gavrilov, Evgenii Sidenko, Hefeng Dong
  Chairperson(s): Hefeng Dong, Evgenii Sidenko
• Lecture: Deep Learning Approach to Remove Distributed Acoustic Sensing Instrumental Noise
  Paper ID: 2211
  Author(s): Olivia Collet, Xihao Gu, Roman Isaenkov, Evgenii Sidenko, Pavel Shashkin, Roman Pevzner, Konstantin Tertyshnikov, Boris Gurevich, Christine Erbe, Sasha Gavrilov
  Presenter: Olivia Collet
  Abstract: Several recent studies have demonstrated the potential of distributed acoustic sensing (DAS) for subsea monitoring. Deploying DAS on pre-existing submarine cables provides a means to record, with large spatial and temporal coverage, a broad range of acoustic signals produced by various underwater sound sources such as marine fauna and anthropogenic activities. Yet, DAS typically exhibits self-noise of varying intensity, part of which comprises instrument-specific noise generated by the DAS interrogator unit. Laboratory studies have shown that the longer the cable, the higher the noise floor is. To mitigate this noise, we have explored the use of machine learning and trained a neural network to establish a mapping between an input and a target image. The target images consist of clean seismic synthetic gathers, while the input images are made up of these same gathers with added instrumental noise recorded on an acoustically isolated coil in the laboratory. Once trained, the neural network can be applied to any DAS data acquired with the same type of interrogator and similar acquisition settings as the ones used for the instrumental noise records. This method has proven effective on seismic field datasets acquired with downhole and surface DAS. In this work, we assess how this approach can help enhance acoustic signals recorded with DAS on subsea telecommunication cables.
• Corresponding author: Dr Olivia Collet
  Affiliation: Center for Exploration Geophysics, Curtin University
  Country: Australia