Subsurface Feature Modeling With Probabilistic Position-Based Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing approaches for identifying subsurface features in a subsurface volume of interest struggle with handling large volumes of geophysical data, requiring subjective analysis or generating binary results without providing uncertainty or probabilistic nature, and are narrowly tailored.
Innovation Solution
A computer-implemented method using machine learning techniques to train a model with multiple tiers of categorization, generating objective and interpretable estimates of subsurface features, including uncertainty as probability, based on subsurface data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional methods are used for identifying subsurface features, then the process can be performed with simple tools, but the analysis becomes subjective and generates binary results without uncertainty information
Solution Approach 1:
The patent replaces traditional mechanical/manual analysis methods with machine learning-based automated analysis. The system uses trained models to process geophysical data and generate probabilistic identification results, eliminating subjective human analysis while providing reliability through algorithmic consistency and uncertainty quantification.
Solution Approach 2:
The patent introduces machine learning models as intermediaries between raw geophysical data and subsurface feature identification. These models act as mediators that process large volumes of data, apply learned patterns, and generate objective probabilistic results, bridging the gap between data and interpretation while providing uncertainty information.
2Reliability
If machine learning models are trained on large volumes of geophysical data, then objective and probabilistic identification is achieved, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on large volumes of geophysical data before actual subsurface feature identification. The models are trained offline on extensive datasets to learn patterns and relationships, so that during actual operation, they can quickly process new data and provide probabilistic results without requiring real-time analysis of massive datasets.
3Loss of information
If traditional binary classification is used, then the method is simple to implement, but it does not provide uncertainty or probabilistic information about the identification
Solution Approach 1:
The patent changes the output parameter from binary classification to probabilistic classification. Instead of simply identifying whether a feature is present or absent, the machine learning models output probability values that indicate the confidence or uncertainty of the identification. This parameter change provides uncertainty information while maintaining a relatively simple model structure.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, devices, and methods are disclosed for identifying subsurface features as a function of position in a subsurface volume of interest. A computer- implemented method may include obtaining training subsurface data and corresponding training subsurface feature data; obtaining an initial subsurface feature model including tiers of elements; generating a conditioned subsurface feature model by training the initial subsurface feature model using the training subsurface data and the corresponding training subsurface feature data; and storing the conditioned subsurface feature model in the non-transient electronic storage.