Azimuth Angle Dependent Full Waveform Inversion Velocity Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating velocity models in geophysics, such as conventional full waveform inversion, face challenges with local minima and non-linear effects, especially when low-frequency information and long offset data are lacking, leading to inaccurate subsurface structure imaging.
Innovation Solution
The implementation of Azimuth and Angle Dependent Full Waveform Inversion (AADFWI) decomposes source and residual wavefields into subsurface azimuth and reflection angle domains, using optical flow to compute gradient directions and apply angle filtering, which helps recover long wavelength structures and avoid cycle skipping, thereby generating a more accurate velocity model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional full waveform inversion is used to generate velocity models, then the process can be performed with standard data, but the results suffer from local minima and non-linear effects leading to inaccurate imaging
Solution Approach 1:
The wavefield is segmented into different azimuth and reflection angle components, allowing the inversion process to handle different directional information separately. This segmentation enables more stable convergence by processing gradient directions in a structured manner across multiple azimuth-angle domains.
Solution Approach 2:
The method introduces additional dimensions by decomposing the wavefield into azimuth and reflection angle domains. This dimensional expansion transforms the conventional single-domain inversion into a multi-dimensional approach, providing more gradient direction information and improving both accuracy and convergence stability.
2Ease of manufacture
If low-frequency information and long offset data are lacking, then data acquisition is simplified, but the velocity model generation becomes inaccurate due to local minima
Solution Approach 1:
The method changes the parameter space by introducing azimuth and reflection angle domains. This transformation allows the inversion to extract more information from existing data by analyzing gradient directions across multiple parameters, compensating for the lack of low-frequency and long-offset information.
Solution Approach 2:
By adding azimuth and angle dimensions to the analysis, the method creates additional informational pathways that do not depend on traditional low-frequency or long-offset data, thereby maintaining accuracy with simplified acquisition.
3Measurement precision
If azimuth and angle domain decomposition is implemented, then long wavelength structures are recovered and cycle skipping is avoided, but the computational complexity increases
Solution Approach 1:
The computational task is segmented into separate azimuth and angle domain processing steps. By dividing the wavefield decomposition into discrete directional components, the method systematically recovers long wavelength structures while organizing the computational workload into manageable segments.
Solution Approach 2:
The wavefield decomposition into azimuth and angle domains is performed as a preliminary action before the main inversion process. This pre-processing step organizes the data in a way that facilitates more efficient subsequent calculations and helps avoid cycle skipping during inversion.
Data Source
AI summary
The present disclosure describes methods and systems, including computer-implemented methods, computer program products, and computer systems, for generating a velocity model for a subsurface area. One computer-implemented method includes obtaining, at a data processing apparatus, source wavefields and residual wavefields, wherein the source wavefields and the residual wavefields are calculated based on seismic data associated with the subsurface area; decomposing, by the data processing apparatus, the source wavefields and the residual wavefields into a subsurface azimuth and reflection angle domain; calculating, by the data processing apparatus, a plurality of gradient direction components based on the decomposed source wavefields and the residual wavefields; scaling, by the data processing apparatus, the plurality of gradient direction components based on azimuths and angles associated with each of the plurality of gradient direction components; calculating a gradient direction based on the scaled gradient direction components; and generating the velocity model using the gradient direction.