Seismic Image Relational Learning for Hydrocarbon Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for detecting hydrocarbon presence in subsurface structures using seismic images are limited by inaccuracies in seismic processing, noise, and complex interactions, leading to unreliable indicators and misinterpretations, particularly in carbonate reservoirs.

Innovation Solution

Employing unsupervised machine learning, specifically variational relational learning (VRL), to analyze relationships between seismic data sets, such as near-stack and far-stack images, to identify hydrocarbon presence without relying on labeled training data, thereby overcoming the limitations of traditional AVO analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional AVO analysis methods are used to detect hydrocarbon presence, then the detection process is straightforward and computationally simple, but the detection reliability is low due to inaccuracies in seismic processing, noise, and complex interactions

Engineering Contradiction:

Improvehydrocarbon detection reliabilityVSAvoiddetection method complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical AVO analysis methods with machine learning-based relational learning. The system uses neural networks to automatically learn complex patterns and relationships in seismic data, substituting manual interpretation and simple computational methods with intelligent algorithms that can handle noise and complex interactions more effectively.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary machine learning model that acts as a mediator between raw seismic data and hydrocarbon detection results. This intermediary layer processes the complex, noisy seismic data through learned relational patterns before producing the final detection output, thereby improving reliability without requiring direct complex processing of raw data.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If multiple hydrocarbon indicators are used to increase confidence in fluid presence, then the detection reliability improves, but the analysis complexity and time required increase significantly

Engineering Contradiction:

Improvefluid presence confidenceVSAvoidanalysis time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent merges multiple hydrocarbon indicators and seismic data attributes into a unified machine learning model. Instead of analyzing each indicator separately and manually integrating results, the system combines all relevant indicators (AVO, amplitude terminations, anomaly consistency, lateral amplitude contrast, etc.) into a single relational learning framework that processes them simultaneously and produces a consolidated detection result.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent transforms multiple seismic data parameters and indicators into a standardized feature set that the machine learning model can process efficiently. By changing the representation of multiple indicators into unified relational features, the system maintains high detection confidence while reducing the time required for analysis compared to traditional multi-indicator workflows.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If manual interpretation and human intervention are used in seismic data analysis, then the detection accuracy can be maintained through expert judgment, but the productivity and efficiency are reduced

Engineering Contradiction:

Improvehydrocarbon detection efficiencyVSAvoiddetection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements a self-service system where the machine learning model automatically performs the entire hydrocarbon detection process without requiring manual interpretation or human intervention. The relational learning model self-adjusts and learns from the data, automatically identifying patterns and making detections, thereby maintaining high productivity while preserving accuracy through the model's learned expertise.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent incorporates feedback mechanisms where the machine learning model continuously learns from detection results and adjusts its parameters accordingly. This feedback loop allows the system to maintain and improve detection accuracy over time while operating autonomously, eliminating the need for continuous manual oversight while preserving high measurement precision.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12468057B2Detection of hydrocarbon presence in subsurface from seismic images using relational learning

Publication Date: 2025.11.11 EXXONMOBIL TECHNOLOGY & ENGINEERING CO

AI summary

A computer-implemented method for detecting geological elements or fluid in a subsurface from seismic images is disclosed. Seismic data may be analyzed to identify one or both of fluid or geologic elements in the subsurface. As one example, the analysis may include unsupervised learning, such as variational machine learning, in order to learn relationships between different sets of seismic data. For example, variational machine learning may be used to learn relationships among partially-stack images or among pre-stack images in order to detect hydrocarbon presence. In this way, an unsupervised learning framework may be used for learning a Direct Hydrocarbon Indicator (DHI) from seismic images by learning relationships among partially-stack or pre-stack images.