FWI Cycle Skipping via Partial Match Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Full Waveform Inversion (FWI) methods face challenges with cycle skipping and amplitude inconsistencies, leading to convergence issues and reduced model resolution, especially when starting from inaccurate initial velocity models or low-quality seismic data.
Innovation Solution
The use of partial match filtering (PMF) to generate auxiliary data by partially matching recorded and synthetic data, which replaces one of the datasets in the objective function to update the model, thereby avoiding cycle skipping and amplitude inconsistencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional FWI minimizes misfit between recorded and synthetic data, then model convergence is achieved, but cycle skipping occurs when initial model is inaccurate
Solution Approach 1:
The patent introduces an intermediary objective function based on matching filters that acts as a mediator between the recorded and synthetic data. Instead of directly minimizing the misfit between recorded and synthetic data (which causes cycle skipping), the method uses matching filters to create an intermediate representation that aligns the phases of both datasets, thereby avoiding cycle skipping while maintaining convergence reliability
Solution Approach 2:
The patent changes the parameterization of the objective function from direct data misfit to a phase-aligned representation using matching filters. By transforming the objective function to operate on phase-aligned data rather than raw data, the method enables accurate timing measurement even when the initial velocity model is inaccurate, thus preventing cycle skipping
2Loss of information
If least-squares difference is used as objective function, then amplitude information is utilized, but amplitude inconsistencies between recorded and modeled data cause convergence issues
Solution Approach 1:
The patent extracts the phase information from the recorded and synthetic data using matching filters, separating it from the amplitude information. By taking out the phase component and aligning it independently, the method allows amplitude inconsistencies to be handled separately, preventing them from causing convergence issues while still utilizing amplitude information in the updated objective function
Solution Approach 2:
The patent applies partial matching through the matching filter approach, where the objective function focuses on achieving phase alignment rather than complete amplitude matching. This partial action on phase alignment first, followed by amplitude consideration, prevents convergence issues caused by amplitude inconsistencies while still utilizing amplitude information
Data Source
AI summary
Methods and apparatuses for seismic exploration of an underground structure obtain improved images by integrating partial match filtering in an FWI. Filtered (auxiliary) data replaces one of the observed data and the synthetic data in the FWI's objective function to avoid cycle skipping.