Seismic Horizon Extraction Using Contradictory Cluster Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for horizon extraction from seismic data are inefficient, often requiring significant manual intervention and produce inaccurate, contradictory, or self-contradictory results due to reliance on propagation, dip and azimuth estimation, and local orientation, especially in the presence of faults and noise.
Innovation Solution
A method involving divisive hierarchical clustering and minimum cut algorithms to identify and partition contradictory clusters in extrema voxel data, ensuring each cluster represents a single horizon by addressing geologically unfeasible configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If propagation-based horizon extraction is used, then horizon data can be obtained, but the results are inaccurate and contradictory due to reliance on dip and azimuth estimation
Solution Approach 1:
The patent extracts and removes contradictory clusters from the horizon data by identifying geologically unfeasible configurations (e.g., horizons that cross or overlap improperly) and systematically eliminating them. This extraction approach separates valid horizon data from invalid data, resolving the contradiction between obtaining horizon data and ensuring its accuracy and consistency.
2Measurement precision
If manual intervention is used in horizon extraction, then accuracy can be improved, but the process becomes time-consuming and inefficient
Solution Approach 1:
The patent implements an automated system that performs horizon extraction and contradiction detection without requiring manual intervention. The computer-executable instructions automatically identify contradictory clusters, determine their geological feasibility, and remove invalid data, enabling the system to serve itself and achieve both high accuracy and high productivity simultaneously.
Solution Approach 2:
The patent replaces manual mechanical interpretation processes with computational algorithms. Instead of relying on human interpreters to manually identify and correct contradictory horizons, the system uses automated computer-based methods to detect and resolve geological inconsistencies, thereby eliminating the trade-off between manual accuracy and automated speed.
3Reliability
If divisive hierarchical clustering is used to partition clusters, then contradictory clusters can be identified, but computational complexity increases
Solution Approach 1:
The patent applies divisive hierarchical clustering to segment the horizon data into distinct clusters, then systematically partitions these clusters to identify contradictory configurations. By breaking down the complex problem of horizon validation into smaller, manageable cluster partitioning tasks, the method achieves reliable contradiction detection while managing computational complexity through structured segmentation.
Data Source
AI summary
Computer-implemented methods, apparatus, and computer programs disclosed herein are for obtaining horizon data, and comprise determining at least one extrema binary volume from a seismic data volume comprising a plurality of voxels and assigning a predetermined extrema value to each one of the found extreme voxels; determining an extrema graph from the at least one extrema binary volume; determining a neighbour graph from the extrema graphs; partitioning the nodes of the neighbour graph into a set of clusters, wherein a cluster contains a plurality of connected nodes representing extrema voxels of the at least one extrema binary volume, and each node is part of one cluster only; for each subset of the set of clusters, identifying whether or not the subset is a contradictory set of clusters according to at least one first predetermined condition; hierarchically partitioning the neighbour graph into a plurality of subgraphs, for example using a minimum cut approach, each subgraph being provided by a separate non-contradictory cluster, which is not part of any contradictory sets of clusters; and obtaining horizon data representing a plurality of horizons from the non-contradictory clusters inducing the plurality of subgraphs of the neighbour graph.