We introduce an iterative workflow that uses data-driven methods to augment time-domain full waveform inversion (FWI) by predicting low frequency seismic data. The predicted data are used to invert a low wavenumber velocity model. When this low wavenumber model is used as an initial velocity model, it helps to reduce the risk of cycleskipping in FWI. Synthetic tests on both acoustic and elastic data demonstrate that when FWI starts with the updated low wavenumber velocity model, it produces more accurate results compared to an initial model without this update.

Presentation Date: Wednesday, October 14, 2020

Session Start Time: 9:20 AM

Presentation Time: 9:20 AM

Location: Poster Station 3

Presentation Type: Poster

Keywords:
neural network, iteration 1, training data, artificial intelligence, high frequency data, initial model, machine learning, deep learning, exploration geophysicist 10, seismic data
Subjects:
Reservoir Characterization, Information Management and Systems, Seismic processing and interpretation
This content is only available via PDF.
2020. Society of Exploration Geophysicists
You can access this article if you purchase or spend a download.