Neural Network Fault Polygon Extraction for Accurate Seismic Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying faults in subterranean formations are inefficient and lack accuracy, which hinders the exploration and production of hydrocarbons.
Innovation Solution
A method involving the generation of structural attributes such as dip and azimuth from seismic data, followed by input into a neural network to identify faulted and non-faulted areas, and the creation of fault polygons using an unsupervised neural network for precise fault mapping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional methods are used to identify faults in subterranean formations, then the process can be completed, but the efficiency is low and accuracy is insufficient
Solution Approach 1:
The patent replaces traditional mechanical/manual fault identification methods with an unsupervised neural network system. The neural network processes seismic data to automatically detect and characterize faults, achieving both higher efficiency and improved accuracy compared to conventional approaches. The system uses deep learning algorithms to analyze seismic attributes and generate fault polygons automatically.
Solution Approach 2:
The patent transforms fault identification from manual interpretation to automated parameter-based detection. The system extracts multiple seismic attributes (amplitude, frequency, impedance) and feeds them into the neural network as input parameters. The network processes these parameters to output fault location and characterization data, enabling efficient and accurate automated fault detection.
2Measurement precision
If existing workflows are used for fault polygon extraction, then fault locations can be identified, but processing time and financial resources are excessive
Solution Approach 1:
The patent replaces time-consuming traditional workflows with an automated neural network system. The unsupervised neural network processes seismic data through parallel computing operations, significantly reducing processing time while maintaining high accuracy in fault polygon extraction. The system automatically generates fault polygons without requiring manual intervention in the analysis process.
Solution Approach 2:
The neural network system performs self-service processing by automatically learning fault patterns from seismic data without human intervention. The unsupervised learning capability allows the system to independently identify fault characteristics and generate accurate fault polygons, eliminating the need for manual analysis and reducing overall processing time and costs.
Data Source
AI summary
Disclosed are methods, systems, and computer-readable medium to perform operations including: obtaining seismic data describing a subterranean surface; generating, based on the seismic data, structural attributes for the subterranean surface; providing the structural attributes as input to a neural network for identifying faults in the subterranean surface, where an output of the neural network includes faulted areas and fault-free areas in the subterranean surface; and generating one or more fault polygons by bounding the faulted areas in the subterranean surface, where the one or more fault polygons are graphical representations of the faulted areas on a map of the subterranean surface.