Machine Learning Borehole Sonic Data Interpretation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for interpreting borehole sonic dispersion data are computationally expensive, limited in their ability to handle anisotropic formations, and lack automation, failing to accurately capture nonlinear effects and identify anisotropy mechanisms in sonic signals.
Innovation Solution
The use of machine learning-based approaches, specifically training neural networks with generated datasets from mode searches or analytical/numerical methods, to approximate solutions for borehole sonic data inversion, enabling faster and more accurate estimation of model parameters with associated uncertainties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classical model-based inversion methods are used to interpret borehole sonic dispersion data, then formation elastic properties can be obtained, but the computational cost is excessively high and the method cannot accurately handle strongly anisotropic formations
Solution Approach 1:
The patent pre-computes dispersion curves for a comprehensive range of formation parameters including anisotropic conditions before actual field data interpretation. These pre-computed results are stored and used during inversion, eliminating the need for real-time complex calculations while maintaining accuracy for strongly anisotropic formations
Solution Approach 2:
The patent creates a lookup table containing pre-computed dispersion curves that serve as a reference database. During inversion, the method searches this table for matching patterns rather than performing full model-based calculations, significantly reducing computational time while preserving measurement precision
2Measurement precision
If physics-based analytical or numerical solutions are used for forward modeling, then accurate borehole sonic data interpretation can be achieved, but the computational expense prevents usage for time-sensitive tasks
Solution Approach 1:
The patent performs forward modeling calculations in advance for a wide range of physical conditions and stores the results in a pre-computed database. When actual borehole sonic data needs interpretation, the method queries this database rather than performing new forward modeling, achieving both accuracy and speed for time-sensitive applications
Solution Approach 2:
The patent creates a comprehensive lookup table with pre-computed physics-based solutions that can be quickly searched and matched against field data, replacing time-consuming numerical solutions with fast database queries while maintaining interpretation accuracy
3Ease of operation
If signal processing methods are used for labeling and extraction of dispersion modes, then the process can be simplified, but the method completely ignores physics and fails to capture nonlinear effects in strongly anisotropic formations
Solution Approach 1:
The patent uses pre-computed physics-based dispersion curves stored in a lookup table as a reference for labeling and extracting dispersion modes from field data. This approach maintains physical accuracy for identifying nonlinear effects while providing automated, straightforward matching procedures that are easy to operate
Solution Approach 2:
The patent creates a unified lookup table that handles both isotropic and anisotropic formation conditions, as well as multiple dispersion modes, within a single framework. This universal table enables accurate physics-based interpretation across diverse conditions while maintaining operational simplicity through consistent matching procedures
4Productivity
If existing inversion methods are used, then borehole sonic data can be interpreted, but the methods are not fully automated and require parameter tuning for challenging cases
Solution Approach 1:
The patent implements an automated inversion system that uses the pre-computed lookup table to automatically match field data without requiring manual parameter tuning. The method self-adjusts by searching the comprehensive database for the best match, eliminating the need for operator intervention in parameter optimization even for challenging anisotropic cases
Data Source
AI summary
The subject disclosure relates to the interpretation of borehole sonic data using machine learning. In one example of a method in accordance with aspects of the instant disclosure, borehole sonic data is received, and machine learning is used to interpret the borehole sonic data.