Full-waveform inversion with CNN-based velocity representation extension: Field data application
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Full waveform inversion (FWI) updates the velocity model by minimizing the discrepancy between observed and simulated data. However, discretization errors in numerical modeling and incomplete seismic data acquisition can introduce noise,which propagates through the adjoint operator and affects the accuracy of the velocity gradient, thereby impacting the FWI inversion accuracy. To mitigate noise in the gradient, a convolutional neural network (CNN) is utilized to refine the velocity model before performing the forward simulation, aiming to reduce noise and provide a more accurate velocity update direction. Data loss is used simultaneously to update both the velocity and the neural network parameters, thereby forming a selfsupervised learning procedure. We propose two implementation schemes, which differ in whether the velocity update passes through the CNN. Since in both methods, the velocity representation is extended (VRE) by using a neural network (which may act as a filter) in addition to the grid values, we refer to this approach as VRE-FWI. Testing conducted using real data collected in Western Australia demonstrates that the developed VRE-FWI provides higher velocity inversion accuracy compared to traditional FWI at a very small additional cost (˜1% more).
Fields of Study & Subjects
Seismic Imaging and Inversion Techniques
Geophysical and Geoelectrical Methods
Seismic Waves and Analysis
Publication details
- Journal / source
- Second International Meeting for Applied Geoscience & Energy
- Volume
- 1
- Pages
- 1144–1148
- Publication date
- 10 February 2026
- Language
- EN