Neural Operator Well Casing Integrity Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inversion algorithms and beamforming methods fall short in efficiently solving for pipe properties and leak source location parameters, particularly due to bottlenecks in the forward model, even with high-performance computing or GPUs, and often require post-processing that reduces their value for real-time decision-making in oil and gas exploration.
Innovation Solution
The implementation of Fourier Neural Operators (FNO) or Physics-Informed Neural Operators (PINO) to replace traditional numerical forward modeling, allowing for more efficient mapping of material functions to physical responses and vice versa, enabling real-time pipe inspection and leak source location determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional inversion algorithms or beamforming methods are used to solve for pipe properties or leak source location, then the problem can be solved, but the computational efficiency is poor and real-time decision-making is hindered
Solution Approach 1:
The patent replaces traditional numerical forward modeling (mechanical/computational system) with a neural operator that directly maps material functions to physical responses. This substitution eliminates the need for iterative numerical solutions and significantly reduces processing time, enabling real-time pipe inspection and leak detection.
Solution Approach 2:
The patent creates a simplified representation (copy) of the complex physical system through a neural operator that captures the essential mapping relationships between material properties and physical responses. This copy allows for rapid inference without repeatedly solving the full physical model, thereby improving computational efficiency.
2Speed
If traditional forward modeling is used, then physical accuracy can be maintained, but the speed of data processing is slow and real-time applications are limited
Solution Approach 1:
The patent substitutes traditional numerical forward modeling with a neural operator that has been trained to maintain physical accuracy while enabling rapid data processing. The neural operator learns the mapping relationships from training data, allowing for fast inference that preserves physical fidelity without the computational burden of traditional methods.
3Ease of operation
If post-processing is used to obtain inversion or beamforming results, then comprehensive analysis can be performed, but the value for real-time decision-making is reduced
Solution Approach 1:
The patent performs the essential mapping from material functions to physical responses in advance through training the neural operator. During actual application, the pre-trained operator can immediately provide results without requiring time-consuming post-processing, enabling real-time decision-making while maintaining comprehensive analysis capabilities.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly enhances the efficiency of pipe property determination and leak source localization, enabling real-time decision-making by overcoming the limitations of traditional methods and improving the accuracy and speed of data processing.
Implementation Method 1
performing a Fourier Neural Operator (FNO) or Physics-Informed Neural Operator (PINO) inversion to determine the pipe status or leak source location
Implementation Method 2
The implementation of Fourier Neural Operators (FNO) or Physics-Informed Neural Operators (PINO) to replace traditional numerical forward modeling, allowing for more efficient mapping of material functions to physical responses
Data Source
AI summary
A method and non-transitory storage computer-readable medium for performing a neural operator on one or more wellbore measurements. The method may comprise o obtaining one or more measurements, performing a measurement normalization on the one or more measurements to form one or more normalized measurements, forming a material function with the one or more normalized measurements, and forming a neural operator generated physical response with a neural operator and the material function. The method may further comprise forming a beamforming map with the one or more measurements, and forming a neural operator leak source location map with a neural operator and the one or more measurements.