Lithofacies Estimation Using Convolutional Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual analysis of well logs by petrophysicists is time-consuming, costly, and lacks accuracy in estimating lithofacies, leading to variable results depending on the analyst.
Innovation Solution
A method and apparatus using an artificial intelligence model, specifically a convolution neural network (CNN) structure, to estimate lithofacies by learning from well logs, which includes forming a lithofacies estimation model through training with diverse data sets and employing error correction to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual log analysis by domain experts is used, then high accuracy in lithofacies estimation is achieved, but it requires huge efforts, high costs, and time
Solution Approach 1:
The patent replaces the manual mechanical analysis process performed by petrophysicists with an automated deep learning system. The CNN-based lithofacies estimation model automatically processes well log data, substituting human expert analysis with an algorithmic system that achieves comparable or superior accuracy while dramatically reducing time and cost requirements.
Solution Approach 2:
The patent creates a digital model that copies and learns from expert knowledge embedded in training data. By training the neural network on labeled well log data with known lithofacies, the system replicates expert estimation capabilities without requiring actual expert involvement in each analysis case, thereby eliminating the time and cost penalties of manual analysis.
2Measurement precision
If manual log analysis by domain experts is used, then accurate lithofacies estimation is achieved, but high costs are incurred
Solution Approach 1:
The patent replaces expensive manual expert analysis with an automated computational system. Once the deep learning model is trained, it can perform unlimited analyses at minimal marginal cost, eliminating the recurring high costs associated with hiring and retaining domain experts for each well log analysis project.
Solution Approach 2:
The trained lithofacies estimation model becomes a self-sufficient system that automatically processes well log data without requiring ongoing human expert intervention. The model serves itself by making predictions directly from input data, eliminating the need for continuous expert labor and associated costs.
3Adaptability or versatility
If manual log analysis by domain experts is used, then lithofacies estimation is performed, but variable results are obtained depending on who analyzed it
Solution Approach 1:
The patent replaces the variable human judgment process with a deterministic computational system. The deep learning model applies the same learned rules consistently to all inputs, eliminating the subjectivity and variability inherent in different experts' interpretations while maintaining the ability to handle diverse well log scenarios.
Solution Approach 2:
The patent transforms the subjective, variable parameters of human expert judgment into objective, fixed parameters encoded in the neural network weights and biases. These parameters remain constant across different analyses, ensuring reproducible results while the model's architecture allows it to adapt to different data types and scenarios.
4Productivity
If automated deep learning model is used, then analysis time is reduced, but model training requires significant computational resources
Solution Approach 1:
The patent performs the computationally intensive work in advance by training the deep learning model on large datasets before deployment. Once trained, the model requires minimal computational resources during actual lithofacies estimation, shifting the energy burden from the operational phase to the initial training phase, thereby achieving fast analysis speeds with low ongoing energy consumption.
Data Source
AI summary
Disclosed are a method and apparatus for estimating lithofacies by learning well logs. The method includes a model formation step of forming lithofacies estimation model to output lithofacies corresponding to measured depth when the well logs are input based on train data sets including train data having values of multiple factors included in the well logs, the values being arranged corresponding to measured depth, and label data having lithofacies corresponding to measured depth as answers, and lithofacies estimation step of inputting unseen data having values of multiple factors included in well logs acquired from a well at which lithofacies are to be estimated, the values being arranged corresponding to measured depth, to the lithofacies estimation model to estimate lithofacies corresponding to measured depth.