Fault Picking in Seismic Volumes Using Minimum Spanning Tree
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seismic data interpretation methods face challenges in accurately picking faults due to noise, multipath effects, and interference, which reduce the accuracy of seismic traces and make it difficult to identify subtle structural features in 3D seismic volumes.
Innovation Solution
The method employs a minimum spanning tree (MST) process to pick faults by generating a principle grid from seismic attribute data, using seeds to represent locations in the seismic volume, and applying a least costs analysis to define parent-child relationships and confidence levels, thereby interpolating a fault surface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional seismic trace interpretation methods are used, then the process can handle large volumes of seismic data, but the accuracy of fault picking is reduced due to noise, multipath effects, and interference
Solution Approach 1:
The patent extracts and isolates fault-related seismic attributes from the complete seismic trace data. By selecting specific attributes that are most indicative of fault structures and removing unrelated trace components, the method focuses computational effort on the most relevant signals while filtering out noise and interference that would otherwise degrade picking accuracy
Solution Approach 2:
The patent introduces a cost function as an intermediary computational layer between the raw seismic attributes and the final fault picks. This cost function serves as a mediator that integrates multiple seismic attributes, applies weighting based on geological knowledge, and produces a unified confidence metric that guides the picking algorithm, thereby improving accuracy in the presence of noise
2Measurement precision
If manual interpretation of seismic traces is performed, then detailed fault analysis can be conducted, but the time required to interpret thousands or millions of traces increases significantly
Solution Approach 1:
The patent implements an automated fault picking system that performs the interpretation task without human intervention. The algorithm autonomously computes seismic attributes, evaluates cost functions, identifies fault candidates, and generates pick results automatically. This self-service capability eliminates the time-consuming manual review process while maintaining consistent application of picking criteria across all seismic data
Solution Approach 2:
The patent replaces the mechanical manual interpretation process with an automated computational algorithm. Instead of human interpreters visually examining and manually picking faults from seismic traces, the system uses computer-based attribute calculation, cost function evaluation, and automated pick generation, thereby dramatically reducing interpretation time while maintaining or improving accuracy
3Manufacturing precision
If a minimum spanning tree process is applied to pick faults, then the accuracy and resolution of fault surfaces are enhanced, but the computational complexity of the surface construction process increases
Solution Approach 1:
The patent segments the fault surface construction process into distinct computational stages: attribute calculation, cost function evaluation, candidate identification, and pick refinement. By dividing the overall task into smaller, manageable segments, the system achieves high precision through systematic processing while keeping computational complexity controlled through modular implementation of each stage
Data Source
AI summary
Picking a fault in seismic data samples is described. In one example, a minimum spanning tree is used. In another example, input seismic attribute data is determined based on the seismic data samples. Seeds are selected that represent locations in the seismic volume using the attribute data. A principle grid is generated using the seeds. A fault is picked in the seismic volume by applying a least costs process, for example a minimum spanning tree, to the principle grid. The fault is then interpolated to generate a fault surface of the seismic volume.