Label Propagation for Dense Subsurface Surface Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for modeling subsurface volumes in the oil and gas industry are computationally expensive and inefficient, particularly when dealing with large data sets, and often result in sparse or inaccurate representations of geophysical properties, failing to provide dense and adjustable surfaces that effectively span the entire volume.
Innovation Solution
A method involving label propagation techniques is employed to generate a labeled volume by assigning vectors representing dip, azimuth, and confidence to each sample in a data volume, allowing for the propagation of labels to create a dense set of surfaces that span the entire subsurface volume, enabling user interaction and enhanced representation of geologic features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods (phase unwrapping, global optimization, plane-wave destruction) are used to compute subsurface models, then certain aspects of surface extraction can be achieved, but the computational cost becomes excessively high and processing time increases significantly
Solution Approach 1:
The patent segments the label propagation process into multiple passes, where in each pass labels are propagated only to neighboring traces that have not yet been assigned a label. This segmentation of the propagation process into discrete, manageable passes significantly reduces computational complexity compared to global optimization methods, while still achieving complete label assignment across the entire volume.
Solution Approach 2:
The patent performs preliminary actions by first computing a vector volume from the data volume, which provides directional information (dip, azimuth, confidence) that guides subsequent label propagation. This preliminary computation of vector fields enables more efficient and accurate label propagation without requiring iterative global optimization.
2Quantity of substance
If existing methods are used to generate subsurface surfaces, then surfaces can be extracted, but they result in sparse representations with gaps rather than dense surfaces that span the entire volume
Solution Approach 1:
The patent makes the label propagation system universal by allowing labels to be propagated along vector directions that are dynamically determined at each location based on the vector volume. This multi-functional approach enables the same propagation mechanism to handle various geological configurations and produces dense, continuous surfaces throughout the entire volume without requiring separate processing for different regions.
Solution Approach 2:
The patent introduces a new dimension of control by using vector directions (dip, azimuth) to guide label propagation through the volume. Instead of propagating labels along fixed grid directions, the vector-based approach adds directional information as a new dimension, enabling labels to propagate along optimal paths through complex geological structures and resulting in dense, accurate surface representations.
3Reliability
If iterative global optimization methods are used to compute flattened volumes, then surfaces can be determined, but multiple iterations are required for complex data increasing processing time
Solution Approach 1:
The patent substitutes the mechanical iterative optimization process with a more direct vector-based label propagation mechanism. Instead of iteratively adjusting surfaces to minimize error, the system computes vector directions from the data and propagates labels along these vectors in a single pass through multiple trace groups, eliminating the need for time-consuming iterative global optimization while maintaining accuracy.
Data Source
AI summary
Method and system are described for generating a stratigraphic model of a subsurface volume. Measured geophysical data are converted into a vector volume (106) by assigning to each sample in each trace in the data volume a vector representing dip, azimuth or confidence. Then a labeled volume is generated from the vector volume by assigning a label to each sample in an initial trace (1006), then selecting a propagation pattern (1008) and propagating the labels to other traces (1010). Horizons can be extracted (110) from the labeled volume, and utilized to enhance the process of producing hydrocarbons (114).