"Machine learning algorithms for automated seismic ambient noise processing applied to DASacquisition"

Citations Over TimeTop 21% of 2018 papers

Abstract

Distributed acoustic sensing (DAS) is an emerging technology used to record seismic data that employs fiber optic cables as a probing system. Since September 2016, a DAS array has been deployed beneath Stanford campus in the existing fiber optic telecommunication conduits. This technology transforms the existing telecomm infrastructure into a large IoT sensor network recording continuous dense seismic data at low cost, with many potential smart city applications. However, due to data volumes, data acquired in such a manner will go to waste unless we significantly automate the processing workflow. We introduce relevant data features for exploratory data analysis and use unsupervised learning to quickly identify coherent, repeating noises which inhibit reliable extraction of useful signals. We use a convolutional neural network for detecting temporally and spatially transient noise and selectively filter it out of the data in order to generate ambient seismic noise fields that are suitable for interferometry purposes. We then combine a convolutional neural network classifier with a Markov decision process to automatically map the recorded data onto its geometry, which opens up the potential to extend this type of acquisition to other existing fiber optic networks in urban areas. Presentation Date: Start Time: Location: Presentation Type:

Related Papers

• → Ambient Noise(2017)4 cited
• → The influences of large earthquake signals on the recovery of surface waves from ambient noise cross-correlation functions*(2020)2 cited
• → Comparison of wind driven ambient noise in shallow and deep-water ocean environments(2007)2 cited
• → Characterization and utilization of heterogeneous ambient noise field for imaging subsurface structure in the Itoshima Peninsula, Japan(2021)