Multi-Task Neural Network Salt Modeling for Focused Seismic Imaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Salt model building in subsurface hydrocarbon exploration is challenging due to strong diffraction and poor image focus, leading to complex and time-consuming manual interpretation, especially in salt-dominated environments, which complicates the detection of geological structures and hydrocarbon presence.

Innovation Solution

A computer-implemented method using machine learning to generate a salt feature model by training a neural network with multiple output channels for salt features and other correlated geological features, such as TOS, BOS, and P-wave velocity, to improve image quality and automate the interpretation process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual interpretation methods are used for salt model building, then the process can handle complex geological structures, but the time consumption and labor intensity increase significantly

Engineering Contradiction:
Improveinterpretation accuracyVSAvoidtime consumption
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical interpretation processes with an automated neural network system. The neural network is trained to automatically detect salt boundaries, top-of-salt, and bottom-of-salt features from seismic data, substituting human experts' manual picking and interpretation work with machine learning algorithms that process data automatically and consistently.

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

Solution Approach 2:

The system enables self-service by allowing the neural network to autonomously perform salt model building without requiring continuous human intervention. The network learns from training data and independently identifies geological features, making the process self-sufficient while maintaining high accuracy through automated feature detection and boundary identification.

Inventive Principle:
Principle #25Self-service

2Device complexity

If traditional single-task neural networks are used, then the model structure is simpler, but the training data utilization is insufficient

Engineering Contradiction:
Improvemodel structureVSAvoidtraining data
Core Design Contradiction:
Device complexityVSQuantity of substance

Solution Approach 1:

The patent implements a multi-task neural network that performs multiple functions simultaneously: detecting top-of-salt, bottom-of-salt, salt boundaries, and other geological features in a single unified model. This multi-functional approach allows the network to leverage diverse training data for multiple objectives, improving data utilization while maintaining a cohesive model structure that handles various detection tasks together.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If human experts perform salt model building, then the interpretation quality is high, but the process becomes labor intensive and ergonomically challenging

Engineering Contradiction:
Improveinterpretation qualityVSAvoidlabor intensity
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent replaces human experts' manual interpretation work with automated neural network processing. The system maintains high interpretation quality by training the network on expert-labeled data, while eliminating the labor-intensive aspects of manual picking, tracking, and boundary identification, thereby improving ease of operation without sacrificing reliability.

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

4Measurement precision

If iterative manual procedures are used for salt model building, then the model accuracy can be improved, but the process takes several months to a year

Engineering Contradiction:
Improvemodel accuracyVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent applies preliminary action by pre-training the neural network on extensive training data before deployment. The network learns from labeled examples during an offline training phase, preparing it to perform accurate salt model building rapidly when deployed. This preliminary training eliminates the need for time-consuming iterative manual refinement during actual model building operations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces the iterative manual refinement process with automated neural network processing. The network rapidly processes seismic data and generates accurate salt models in minutes or hours rather than months, substituting the slow iterative human review and adjustment cycles with efficient machine learning inference while maintaining high model accuracy through the network's learned capabilities.

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

Data Source

PatentEP4334758B1Multi-task neural network for salt model building
Publication Date: 2025.08.20 EXXONMOBIL TECHNOLOGY & ENGINEERING CO
  • EP4334758B1 patent drawingFigure 1
  • EP4334758B1 patent drawingFigure 2A
  • EP4334758B1 patent drawingFigure 2B

AI summary

A method and a system for a multi-task neural network for salt model building is disclosed. Imaging salt in the subsurface may be challenging because salt may be associated with strong diffraction and poor focused image, thereby making it difficult to interpret sediments underneath salt body or near salt flanks. To better image salt in the subsurface, the method and system trains, in combination, multiple aspects related to the subsurface, one of which is the target salt feature, in order to generate a salt feature model. The multiple aspects may include the target salt feature, such as the predicted salt mask, and at least one other salt feature, and one or more subsurface features, such as reconstruction of the input image and P-wave velocity. Thus, the salt model may better image salt, thereby making the seismic migration image more focused and easier to identify geological structures.