Formation Pore and Grain Classification Using 3D Image Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual classification of pore and grain types in formation samples by geologists is laborious, inconsistent, and prone to inaccuracies, especially for 3D volumes, affecting the reliability of wellbore evaluation for oil and gas extraction.
Innovation Solution
A method and system using geometric feature computation and machine learning models to automatically classify pore and grain types in formation samples, leveraging CT images and combining unsupervised and supervised learning techniques with manual data for accurate and consistent classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual classification by geologists is used, then classification can be performed with existing expertise, but the process becomes laborious and inconsistent
Solution Approach 1:
The patent replaces the manual mechanical classification process performed by geologists with an automated image processing and machine learning system. The system uses geometric feature computation and classification algorithms to automatically identify and categorize pores and grains, eliminating the need for manual visual inspection while improving consistency and reducing time requirements.
Solution Approach 2:
The classification system performs self-service by automatically analyzing formation sample images without requiring continuous human intervention. The machine learning model processes images independently, extracting geometric features and assigning classifications autonomously, thereby reducing laborious manual work while maintaining high classification consistency.
2Measurement precision
If manual classification is performed for whole 3D volumes, then complete coverage is achieved, but the complexity becomes extremely complicated
Solution Approach 1:
The patent applies segmentation by dividing the complex 3D volume classification task into manageable components. The system processes images at different scales (thin sections and whole volumes), separates feature extraction into geometric computations, and breaks down classification into structured categories (pore types, grain types), making the overall process more tractable while maintaining precision.
Solution Approach 2:
The patent leverages dimensionality by transitioning from 2D thin section images to 3D whole volume analysis. The system computes geometric features in three dimensions, enabling comprehensive classification of formation samples while managing complexity through automated processing algorithms that scale effectively with dimensionality.
3Adaptability or versatility
If qualitative evaluation by individual geologists is used, then expert judgment is applied, but inconsistencies between geologists arise
Solution Approach 1:
The patent transforms the subjective qualitative evaluation process into objective parameter-based classification. By computing specific geometric features (area, perimeter, shape factors, spatial relationships) and using these as classification parameters, the system maintains flexibility in handling diverse formation samples while ensuring consistent and reliable results across different cases.
Data Source
AI summary
A method is provided for automatically classifying grains, pores, or both of a formation sample. The method includes receiving a digital image representation of the formation sample, and identifying a plurality of pores, grains, or both in the digital image representation. The method also includes computing a plurality of geometric features associated with the pores, grains, or both in the digital image representation, and inputting the geometric features into an unsupervised machine learning model. The unsupervised machine learning model determines a label for each identified pore and grain, the label being a pore-type or a grain-type, and the plurality of geometric features and the labels determined for each pore, grain, or both, are input into a supervised machine learning model. The supervised machine learning model determines a final classification of a pore-type for each pore and a grain-type for each grain in the digital image representation of the formation sample.