Seismic Facies Identification via Machine Learning Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for identifying seismic facies from seismic data are time-consuming and inefficient, often failing to accurately represent the true geometry and connectivity of geological features, leading to bottlenecks in subsurface characterization and hydrocarbon reservoir exploration.

Innovation Solution

The use of machine learning techniques, including unsupervised and supervised learning methods, to classify seismic data into distinct groups based on geometric and textural characteristics, generating 3-D digital images of seismic facies, which leverages computational power to uncover patterns difficult for human interpreters to spot.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If trace-by-trace classification methods are used, then seismic facies can be identified, but the process is time-consuming and creates bottlenecks in subsurface characterization

Engineering Contradiction:
Improveseismic facies identification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the seismic data processing into distinct operational phases: initial trace-by-trace classification to identify candidate regions, followed by targeted analysis of those specific regions. This segmentation allows the time-intensive classification to be focused only where needed, rather than processing entire datasets uniformly, thus reducing overall processing time while maintaining identification accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by performing detailed facies classification only on selected traces or trace groups that meet specific criteria, rather than processing every trace in the dataset. This selective approach processes only the necessary portion of data at high detail, significantly reducing computation time while still achieving accurate facies identification in the most critical areas.

Inventive Principle:
Principle #16Partial or excessive action

2Measurement precision

If trace-by-trace classification is performed, then seismic facies can be identified, but the true geometry and connectivity of geological features are not accurately represented

Engineering Contradiction:
Improveseismic facies identification accuracyVSAvoidgeological feature geometry
Core Design Contradiction:
Measurement precisionVSShape

Solution Approach 1:

The patent merges multiple traces that exhibit similar facies characteristics into continuous facies bodies or zones. By combining adjacent traces with matching classification results, the method reconstructs the spatial continuity and geometric integrity of geological features, accurately representing their true shape and connectivity rather than treating each trace as an isolated unit.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent transitions from analyzing seismic data in a one-dimensional trace-by-trace manner to a multi-dimensional volumetric approach. By integrating information across multiple traces and depth levels, the method reconstructs three-dimensional geological features, preserving their spatial geometry and connectivity in the resulting facies models.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Productivity

If machine learning techniques are used to classify seismic data, then processing efficiency is improved, but computational complexity increases

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent performs preliminary preprocessing of seismic data before applying machine learning algorithms, including data normalization, feature extraction, and initial filtering. This preliminary action prepares the data in an optimized format that reduces the computational burden on subsequent machine learning processing, improving overall efficiency while managing complexity through staged processing.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3526629B1System and method for seismic facies identification using machine learning
Publication Date: 2020.11.25 CHEVRON USA INC
  • EP3526629B1 patent drawingFigure 1
  • EP3526629B1 patent drawingFigure 2
  • EP3526629B1 patent drawingFigure 3

AI summary

A method is described for seismic facies identification including receiving a seismic dataset representative of a subsurface volume of interest; performing a machine learning algorithm on the seismic dataset to identify seismic facies and generate a classified seismic image; and identifying geologic features based on the classified seismic image. The method may be executed by a computer system.