Multi-scale deep network for seismic fault detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for detecting faults in subterranean formations during oilfield operations rely heavily on manual seismic interpretation, which is subjective and inefficient, especially when identifying unknown faults in target seismic volumes.

Innovation Solution

A multi-scale deep network system is employed, where a machine learning model is trained using a training seismic volume with known faults, divided into patches and areas, to generate labels. This model is then applied to a target seismic volume using a sliding window approach to predict fault locations, aggregating results to identify unknown faults.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual seismic interpretation is used for fault detection, then flexibility in analysis is maintained, but detection accuracy and objectivity deteriorate due to subjectivity

Engineering Contradiction:
Improvefault detection accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces manual mechanical interpretation processes with an automated machine learning system. The neural network model automatically analyzes seismic volumes to detect faults, substituting human expert manual analysis with an computational system that provides consistent, objective results without subjective bias while maintaining high detection accuracy.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If manual seismic interpretation is used, then adaptability to different cases is maintained, but processing efficiency and productivity deteriorate

Engineering Contradiction:
Improvefault detection efficiencyVSAvoidanalysis time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent implements preliminary action by pre-training the neural network model on labeled seismic data containing known faults. This pre-training phase enables the model to learn fault patterns and characteristics in advance, so that when applied to target seismic volumes, the model can rapidly detect faults without requiring time-consuming manual analysis, thus significantly improving productivity while reducing analysis time.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If traditional fault detection methods are used, then simplicity of operation is maintained, but detection reliability deteriorates due to subjectivity

Engineering Contradiction:
Improvefault detection reliabilityVSAvoiddetection system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements self-service through the machine learning model that automatically performs fault detection without requiring continuous human intervention or subjective judgment. The model independently analyzes seismic data, identifies fault patterns, and generates detection results consistently, thereby improving reliability by eliminating subjectivity while the automated nature maintains operational simplicity despite the underlying system complexity.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3494284B1Multi-scale deep network for fault detection
Publication Date: 2023.11.08 SERVICES PETROLIERS SCHLUMBERGER SA
  • EP3494284B1 patent drawingFigure 1.1
  • EP3494284B1 patent drawingFigure 1.2
  • EP3494284B1 patent drawingFigure 2

AI summary

A method for detecting an unknown fault in a target seismic volume. The method includes generating a number of patches from a training seismic volume that is separate from the target seismic volume, where a patch includes a set of training areas, generating a label for assigning to the patch, where the label represents a subset, of the set of training areas, intersected by an known fault specified by a user in the training seismic volume, training, during a training phase and based at least on the label and the training seismic volume, a machine learning model, and generating, by applying the machine learning model to the target seismic volume during a prediction phase subsequent to the training phase, a result to identify the unknown fault in the target seismic volume.