Full text loading...
-
Deep-Learning-based Surface Waves Inversion: Application to Pseudo-Shot Gathers from Passive Seismic Data
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 6th Asia Pacific Meeting on Near Surface Geoscience and Engineering, May 2024, Volume 2024, p.1 - 5
Abstract
Summary
Seismic interferometry utilizes cross-correlation between responses from two receivers to create a pseudo-shot gather, treating one as a virtual source. Prioritizing surface wave analysis over body waves due to signal-to-noise ratio challenges, we apply multichannel analysis of surface waves (MASW). However, manual dispersion curve picking in MASW poses challenges. To address this, we can employ a convolutional neural network (CNN) model, eliminating the need for manual picking and utilizing a customized loss function and post-processing neural network. This model successfully estimates S-wave velocity from dispersion images, overcoming challenges of MASW. In this study, we assess the feasibility of the CNN-based model on pseudo-shot gathers generated via seismic interferometry. Acquiring both active shot gathers and ambient noise responses in two distinct fields, we opted for cross-coherence instead of cross-correlation to generate the pseudo-shot gather. For both active shot and pseudo-shot gathers, dispersion images were input into the trained deep learning model, predicting S-wave velocities. Notably, we observed consistent S-wave velocity estimates between active and passive seismic data, indicating the model’s proficiency with passive seismic data even though the CNN model was derived solely from active seismic data.
© EAGE Publications BV