RBF Interpolation for NMR Capillary Pressure Prediction in Carbonates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current NMR methods struggle to accurately determine capillary pressure curves in carbonate formations with complex pore systems, as existing methods are not reliable for predicting reservoir quality due to the variability in pore types and diagenesis processes in carbonates, limiting the conversion of NMR measurements to capillary pressure curves.
Innovation Solution
The use of a radial basis function (RBF) interpolation method with a database of NMR measurements from carbonate samples to predict capillary pressure curves, employing Thomeer Hyperbola parameterization and RBF interpolation to construct a non-linear mapping from NMR spin-echo signals to capillary pressure parameters, allowing for accurate estimation of formation properties like permeability and porosity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional NMR methods are used to determine capillary pressure curves in carbonate formations, then the measurement can be performed in-situ in a borehole, but the prediction accuracy of reservoir quality is poor due to complex pore systems and diagenesis
Solution Approach 1:
The patent introduces an intermediary approach by using a database of core analysis data and machine learning algorithms as a mediator between NMR measurements and capillary pressure curve prediction. The database serves as a knowledge base that captures the complex relationships between NMR signals and reservoir properties, while the machine learning model acts as a translator that converts NMR data into accurate predictions despite the complexity of carbonate pore systems and diagenesis.
Solution Approach 2:
The patent applies parameter changes by transforming the NMR signal characteristics into different parameter spaces through machine learning processing. The model maps NMR relaxation times, amplitudes, and other signal parameters into transformed parameters that directly correlate with capillary pressure characteristics, effectively simplifying the complex relationship between NMR measurements and reservoir quality parameters.
2Reliability
If NMR measurements are converted to capillary pressure curves using traditional methods, then continuous in-situ reading is achieved, but the reliability of the conversion is limited by the variability in pore types and connectivity
Solution Approach 1:
The patent implements feedback mechanisms through iterative machine learning training where the model continuously refines its predictions based on feedback from core analysis data. The system learns from the variability in pore types and connectivity patterns across different carbonate samples, adjusting its internal parameters to improve conversion reliability while maintaining adaptability to diverse pore systems.
Solution Approach 2:
The patent applies preliminary action by pre-processing and analyzing core data from multiple carbonate formations to build a comprehensive database before applying the NMR measurement. This preliminary database construction captures the full range of pore type variations and connectivity patterns, enabling the machine learning model to handle the adaptability challenges during actual NMR conversion with improved reliability.
3Measurement precision
If laboratory core analysis is performed to obtain pore-throat distribution and pore-size distribution, then accurate reservoir characterization is achieved, but the measurements are destructive and have limited range
Solution Approach 1:
The patent uses the machine learning model as a virtual copy of the laboratory core analysis process. Instead of physically analyzing core samples in the laboratory, the NMR measurements serve as a proxy that the machine learning model translates into equivalent pore-throat distribution and pore-size distribution data. This copying approach maintains the accuracy of reservoir characterization while eliminating the destructive nature and limited range of physical core analysis.
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 enables reliable prediction of capillary pressure curves and formation properties, improving the accuracy of reservoir quality assessment in carbonate formations by overcoming the limitations of traditional methods, particularly in complex carbonate rocks.
Implementation Method 1
using a nuclear magnetic resonance (NMR) sensor assembly conveyed in a borehole in the earth formation and obtaining nuclear magnetic resonance (NMR) spin-echo signals indicative of the property of the earth formation
Data Source
AI summary
An apparatus and method for determining a property of an earth formation using a radial basis function derived from a catalog of rock samples. Parameters of a Thomeer capillary pressure fitting curve are derived and used for analyzing rocks with unimodal or multimodal pore size distributions.