Seismic Velocity Model Inversion with Accuracy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic surveying methods struggle to accurately determine velocity models in subterranean formations, particularly due to the neglect of statistical uncertainties in travel time data, leading to low-accuracy velocity models and increased computational complexity.
Innovation Solution
The method involves determining one-dimensional (1D) velocity profiles by inverting observed first-break travel times from seismic traces, using an objective function that minimizes data misfit and incorporates regularization terms to account for uncertainties, thereby generating pseudo-3D models for improved seismic imaging.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional seismic surveying methods are used to determine velocity models, then the processing can be completed with simpler algorithms, but the accuracy of velocity models is low due to neglect of statistical uncertainties in travel time data
Solution Approach 1:
The patent transforms the velocity model determination from a deterministic process to a statistical inversion process by incorporating data covariance matrices that represent uncertainties in travel time measurements. This parameter change allows the algorithm to account for statistical variations in the data, improving velocity model accuracy while maintaining computational tractability through efficient matrix operations.
Solution Approach 2:
The patent replaces traditional mechanical/deterministic inversion algorithms with a statistical mechanics-based approach using covariance matrices and probabilistic frameworks. This substitution enables the system to handle uncertainties in a systematic way, improving measurement precision through statistical optimization rather than simple deterministic calculations.
2Measurement precision
If 3D velocity models are generated to improve imaging resolution, then the seismic image quality improves, but the computational complexity increases drastically
Solution Approach 1:
The patent segments the complex 3D velocity model determination into multiple 1D vertical velocity profiles, each determined independently through statistical inversion. This segmentation reduces the computational burden by breaking down the large-scale 3D inversion problem into smaller, more manageable 1D problems that can be solved efficiently and then assembled into the final 3D model.
Solution Approach 2:
The patent transitions from directly solving a complex 3D inversion problem to solving multiple 1D inversion problems vertically, then combining them. This dimensionality reduction approach maintains the essential vertical velocity variations needed for accurate imaging while avoiding the computational explosion of full 3D statistical inversion.
3Measurement precision
If statistical uncertainties are incorporated into the inversion process to improve velocity model accuracy, then the measurement precision improves, but the computational demands increase
Solution Approach 1:
The patent applies statistical inversion with covariance matrices selectively to the most critical parameters (vertical velocity profiles) rather than attempting full statistical inversion of all 3D velocity parameters. This partial application of statistical methods achieves significant accuracy improvements in the most important velocity components while avoiding the prohibitive computational costs of complete statistical 3D inversion.
Data Source
AI summary
A method, system, and non-transitory computer readable media for seismic imaging of a subterranean formation. The operations include receiving data representing seismic traces corresponding to seismic waves propagating in the subterranean formation, and determining, from the seismic traces, travel time data and a data covariance matrix. The operations include determining a prior velocity model and a prior model covariance matrix for common midpoint locations and generating an objective function based on the travel time data, the data covariance matrix, the prior velocity model, and the prior model covariance matrix. The operations include determining, by minimizing the objective function, values for one-dimensional velocity models and a set of accuracy values. The operations include generating a pseudo-3D model and a set of accuracy values for the pseudo-3D model. Based on the pseudo-3D model and the set of accuracy values for the pseudo-3D model, a seismic image representing the subterranean formation is generated.