This webinar will feature the following four presentations:
Dr. Whitney Trainor-Guitton (Colorado School of Mines), “Machine learning is driving a revolution in seismic ‘listening’ for geoscience”, I will describe our working group’s synthesis on how DAS combined with machine learning can lead to transformative subsurface monitoring. A snapshot of previous work and outstanding challenges will be presented.
Dr. Martijn van den Ende (Université Côte d’Azur), “A Self-Supervised Deep Learning Approach for Blind Denoising of Distributed Acoustic Sensing Data”, The spatial density of DAS measurements allows us to use the full wavefield in our signal processing workflows, rather than treating each sensor separately. In this talk I will detail a simple self-supervised Deep Learning technique that leverages the full wavefield to separate earthquake signals from noise.
Dr. Eileen Martin (Virginia Tech), "Noise Exploration and Detection," DAS enables us to easily collect more data than we could budget time to manually explore, especially in populated areas or around infrastructure. This talk will show simple examples of how to use machine learning to explore and target noise for removal in two DAS datasets.
Fantine Huot (Stanford), “A deep learning model for microseismic detection”, Downhole DAS offers a great opportunity to acquire high-resolution microseismic signals for downhole operation monitoring, but the large volume of continuous data acquired from the DAS fiber also poses a huge challenge for data processing and microseismic event detection. In this talk we will demonstrate a CNN-based deep learning model that tackles this problem with super-human accuracy on a large-volume downhole DAS dataset.