Velocity Model Construction via Hierarchical Basis Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic imaging techniques, such as tomography and full waveform inversion (FWI), are inadequate for accurately modeling seismic wave propagation in complex subsurface regions with varying geological features, leading to incomplete or inaccurate seismic images.
Innovation Solution
Targeted velocity model construction, which involves decomposing the subsurface into hierarchical-scale spatially localized velocity perturbation basis functions, iteratively updating the velocity model to maximize stack power, and using high-quality migration algorithms to refine the seismic image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional tomography or full waveform inversion (FWI) methods are used to model seismic wave propagation, then the processing speed is relatively fast or the method is general, but the accuracy is insufficient in complex subsurface regions with varying geological features
Solution Approach 1:
The patent segments the velocity model into hierarchical-scale spatially localized basis functions, dividing the complex subsurface into manageable components that can be independently optimized. This segmentation allows the method to handle complex geological features by treating different spatial scales and locations separately, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent applies local quality by using spatially localized velocity perturbation basis functions that adapt to specific regional characteristics. Each basis function is optimized for its local geological context, allowing the velocity model to accurately represent varying geological features at different locations while maintaining computational tractability.
2Measurement precision
If the velocity model is updated using tomographic back-projection of error measures, then the method is computationally fast, but the accuracy is highly-approximate and relies on simplified models that only work when the true model is sufficiently smooth
Solution Approach 1:
The patent changes the parameters of the velocity model by decomposing it into hierarchical-scale basis functions with variable spatial frequencies. This parameter transformation allows the model to represent complex, non-smooth geological features that traditional tomography cannot capture, improving both accuracy and reliability simultaneously.
Solution Approach 2:
The patent introduces dynamics by iteratively updating the velocity model through multiple passes, where the basis function coefficients are adjusted based on migration residuals. This dynamic adaptation allows the model to progressively improve accuracy while maintaining robustness through the hierarchical structure.
3Measurement precision
If full waveform inversion (FWI) is used to build the sound speed model iteratively by waveform fitting, then the method is more general and faithful to physics, but it does not necessarily produce the best model for seismic migration and is insufficient at deeper depths
Solution Approach 1:
The patent uses an intermediate representation through spatially localized basis functions that bridge the gap between FWI's physical fidelity and migration quality. These basis functions serve as a mediator that translates waveform fitting results into a velocity model optimized for seismic migration, resolving the contradiction between general applicability and migration-specific performance.
Solution Approach 2:
The patent adds a new dimension to the velocity model by incorporating hierarchical-scale spatial localization. This dimensional enhancement allows the model to capture both deep subsurface features and local variations simultaneously, improving both sound speed accuracy and migration quality across all depths.
4Productivity
If simple tomography techniques are used, then the processing is computationally fast, but the method is highly-approximate and only works when the region through which waves transit is consistent
Solution Approach 1:
The patent segments the velocity model into hierarchical basis functions that can be processed efficiently while capturing complex features. This segmentation maintains computational speed by organizing the problem into manageable scales, unlike traditional methods that require processing the entire complex model at once.
Solution Approach 2:
The patent transforms the velocity model parameters into a hierarchical basis function representation, which enables efficient computation while accurately representing complex geological variations. This parameter change allows the method to maintain productivity while achieving high accuracy in variable regions.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Estimation of velocity models inclusive of receiving seismic data inclusive of data that corresponds to a seismic image, adding a velocity perturbation to a current velocity model that represents a portion of the subsurface responsible for a distortion in the seismic image to generate a perturbed velocity model, generating an image via seismic migration of the seismic data and the perturbed velocity model, generating and assigning a measure of quality to the image, determining whether the measure of quality assigned to the image is an optimal measure of quality at a particular location of the current velocity model, and updating the current velocity model to generate a revised velocity model utilizing the measure of quality assigned to the image when the measure of quality assigned to the image is determined to be the optimal measure of quality at the particular location of the current velocity model.