Cement Evaluation Using Clinging P Detection and Galaxy Pattern Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating cement in wellbore annuli struggle to accurately detect and differentiate between solid cement, mud-contaminated cement, and gas, particularly at the top of the cement annulus, due to the sporadic nature of the 'clinging P' feature and interference from third-interface echoes.
Innovation Solution
A machine learning-based workflow that combines flexural wave data with ultrasonic pulse-echo data to automate the detection of clinging P arrivals and galaxy patterns, using a convolutional neural network to identify features indicative of contaminated cement and the top of cement, thereby enhancing the accuracy and autonomy of cement evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional ultrasonic imaging methods are used to evaluate cement, then the evaluation process requires expert interpretation and manual analysis, but this leads to lower automation and higher subjectivity in cement condition assessment
Solution Approach 1:
The system enables self-service automation by using machine learning models that automatically detect clinging P arrivals and identify cement conditions without requiring expert operator intervention. The neural network processes the ultrasonic signals autonomously, making the system self-sufficient in its analytical function.
Solution Approach 2:
The patent replaces the mechanical/manual expert interpretation system with an automated computational system. Instead of relying on human experts to manually analyze the ultrasonic images and identify cement conditions, the system uses machine learning algorithms to automatically detect patterns and make classifications.
2Productivity
If manual expert analysis is used to interpret cement conditions, then measurement precision can be maintained, but productivity decreases due to time-consuming manual processing
Solution Approach 1:
The system incorporates feedback mechanisms where the machine learning model continuously learns from and refines its detection based on the ultrasonic signal patterns. The model processes the signals in real-time, providing immediate feedback on cement conditions detected, thereby maintaining high accuracy while significantly improving processing speed.
Solution Approach 2:
The patent substitutes manual expert analysis with automated machine learning processing that operates at the speed of computation. The neural network can rapidly analyze large amounts of ultrasonic data, providing both high productivity through fast processing and high measurement precision through sophisticated pattern recognition algorithms.
3Reliability
If the clinging P feature is used to detect contaminated cement, then detection capability improves, but the sporadic nature of the feature reduces reliability
Solution Approach 1:
The patent replaces manual detection methods with automated machine learning algorithms that can consistently identify clinging P arrivals despite their sporadic nature. The neural network processes the ultrasonic signals through multiple computational steps, enhancing its ability to reliably detect and measure these intermittent features through pattern recognition and feature extraction.
Solution Approach 2:
The system uses feedback mechanisms where the machine learning model continuously refines its detection accuracy by learning from the sporadic appearing clinging P features. The model adapts to the intermittent nature of these features through iterative training and validation, improving its reliability in detecting contaminated cement even when the clinging P arrivals are not consistently present.
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 the quantitative and autonomous detection of contaminated cement and the top of cement, improving the precision of well integrity assessment by distinguishing between different cement conditions and reducing reliance on expert intervention.
Implementation Method 1
lowering an ultrasonic tool that implements imaging through steel casing based on the casing (quasi-Lamb) modes including a thickness-mode-dominated measurement through a pulse-echo modality, and a flexural-mode-dominated measurement through a pitch-catch modality
Implementation Method 2
the dispersive flexural mode data yields a modal attenuation of the casing-propagating signal across two receivers
Implementation Method 3
When the annular fill is comprised of a solid with a compressional wave velocity that intersects the dispersive flexural mode phase velocity curve within the signal frequency bandwidth, phase matching to a headwave in the annulus occurs and an additional contribution that follows closely the arrival from within the casing is observed: a feature referred to as a 'clinging P'
Data Source
AI summary
Cement in a wellbore is evaluated by using cement evaluation tools to obtain ultrasonic image information and flexural wave image information. Flexural wave imager waveforms are applied to a trained machine learning system that predicts the presence or lack thereof of a clinging P as a function of depth and azimuth, while the ultrasonic imager impedance maps are applied to a trained machine learning system that identifies galaxy patterns for depths and azimuths, indicating the presence of a third interface echo (TIE) close to the casing. Locations of clinging Ps are compared to locations of galaxy patterns to identify when a TIE could cause a false clinging P determination. Where a disambiguated clinging P is found, contaminated cement is identified that may also locate the top of the cement in the annulus.