Autonomous Rock Drill Cuttings Interpretation via ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional human interpretation of rock drill cuttings is labor-intensive, prone to bias, and inconsistent, leading to high costs and variable assessments across different wells and geological formations.
Innovation Solution
An autonomous system using machine learning models to preprocess, segment, and predict mineralogical or sedimentological data from rock drill cuttings representations, enabling real-time, consistent, and accurate interpretation without the need for on-site human geologists.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human interpretation of rock drill cuttings is used, then expertise and judgment can be applied, but the process becomes labor-intensive and inconsistent
Solution Approach 1:
The patent replaces the manual mechanical process of human geologists examining and interpreting rock drill cuttings with an automated optical system using digital imaging and machine learning algorithms. The system captures images of cuttings, processes them through trained models, and generates mineralogical and sedimentological interpretations automatically, eliminating labor-intensive manual analysis while maintaining or improving accuracy through consistent application of trained algorithms across all samples.
2Adaptability or versatility
If human interpretation is used, then flexible judgment can be applied, but bias and inconsistency arise across different wells and formations
Solution Approach 1:
The patent transforms the interpretive process from subjective human judgment to objective algorithmic analysis by changing the parameters from human expertise variables to standardized machine learning model parameters. The system uses consistent trained models with fixed parameters that are applied uniformly across all wells and formations, eliminating inter-observer variability and bias while maintaining adaptability through the model's ability to recognize diverse rock types and geological contexts.
3Measurement precision
If automated machine learning interpretation is implemented, then consistency and accuracy improve, but system complexity increases
Solution Approach 1:
The patent uses digital copies and representations of physical rock drill cuttings through high-resolution imaging, replacing the need for physical sample handling and manual examination. The machine learning models process these digital copies to generate interpretations, simplifying the overall system by eliminating complex physical manipulation steps while maintaining interpretive accuracy through sophisticated digital analysis.
4Device complexity
If manual processing of rock drill cuttings is used, then simple equipment is required, but time consumption increases significantly
Solution Approach 1:
The patent implements continuous automated processing where the machine learning system operates without interruption to analyze rock drill cuttings as they are collected, eliminating the discontinuous nature of manual examination. The system processes images and generates interpretations in continuous workflow, dramatically reducing total interpretation time while requiring only moderate equipment complexity consisting of imaging devices and computing resources.
Data Source
AI summary
A computer-implemented method that autonomously performs rock drill cuttings interpretation is described herein. The method includes obtaining rock drill cuttings representations. The method also includes preprocessing the rock drill cuttings representations. The method also includes performing unsupervised image segmentation in order to obtain masked representations of such images discriminating rock types. The method also includes performing supervised learning through a custom Convolutional Neuronal Network using the segmented pictures as inputs and a continuous or discrete mineralogical or sedimentological variable of interest as the output. Additionally, the method includes autonomously predicting such mineralogical or sedimentological quantity from new rock drill cuttings pictures using the parameters of the unsupervised segmentation and the trained supervised model created for this purpose.