Multiple-Cased Well Annuli: Sonic Machine Learning for Cement Integrity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current well logging methods are inadequate for diagnosing cement integrity in multiple-cased oil and gas wells, as they are limited to the annulus behind the innermost casing and cannot effectively probe deeper annuli, leading to costly operations and inefficiencies in maintaining well integrity.

Innovation Solution

A method using array sonic data from monopole, dipole, and quadrupole modalities processed via transforms like slowness time coherence (STC) and Radon transforms, followed by machine learning techniques such as support vector machines (SVM), auto-encoders, or convolutional neural networks (CNN) to analyze the status of annuli behind multiple casings, providing a comprehensive evaluation of cement integrity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If current acoustic measurement methods (CBL-VDL, ultrasonic) are used, then single casing string diagnosis is achieved, but multiple casing string diagnosis capability is lost

Engineering Contradiction:
Improvecement integrity diagnosis capabilityVSAvoidapplicability to multiple casing strings
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The acoustic measurement system is designed to perform multiple functions: it can diagnose both single and multiple casing strings using the same tool and measurement methodology. The system processes acoustic signals to evaluate cement integrity in the first annulus and subsequent annuli between multiple casing strings, making the measurement system universally applicable across different well configurations without requiring separate specialized tools for each scenario.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Loss of information

If model-based inversion is applied to invert all parameters, then comprehensive parameter extraction is achieved, but computational feasibility deteriorates for continuous logs covering thousands of feet

Engineering Contradiction:
Improvecompleteness of parameter extractionVSAvoidcomputational efficiency for continuous logs
Core Design Contradiction:
Loss of informationVSProductivity

Solution Approach 1:

The system extracts only the essential and most relevant parameters needed for cement integrity assessment from the acoustic measurements, rather than attempting to invert all possible parameters. This selective extraction approach maintains sufficient diagnostic information while dramatically reducing computational complexity, making it feasible to process continuous logs covering thousands of feet of wellbore within practical time and resource constraints.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of performing complete model-based inversion of all parameters throughout the entire wellbore, the system applies partial inversion focused on key parameters at critical locations. This partial action approach provides sufficient diagnostic capability for cement integrity assessment while avoiding the prohibitive computational burden of exhaustive parameter inversion across all depth intervals.

Inventive Principle:
Principle #16Partial or excessive action

3Adaptability or versatility

If additional measurements and processing approaches are implemented to probe deeper than the first casing, then diagnosis capability beyond first annulus is improved, but measurement physics complexity increases

Engineering Contradiction:
Improvedepth of diagnosis capabilityVSAvoidcomplexity of measurement physics
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The acoustic measurement system segments the complex multiple-casing diagnosis problem into distinct analysis stages: first evaluating the innermost casing and annulus, then sequentially evaluating subsequent casings and annuli. This segmentation allows the system to handle each casing string individually using established acoustic principles, avoiding the need to simultaneously model all casings together, thereby reducing overall measurement physics complexity while maintaining deep diagnosis capability.

Inventive Principle:
Principle #1Segmentation

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

Enables reliable diagnosis of cement integrity beyond the first casing and annulus, facilitating informed decision-making on remedial actions and improving the efficiency and accuracy of well maintenance.

Implementation Method 1

The array sonic data comprising one or more of monopole, dipole and quadrupole modalities from one or more sources

Methodology Applied
Scientific EffectAcoustic wave propagation: Sound

Implementation Method 2

processed via transforms like slowness time coherence (STC) and Radon transforms

Methodology Applied
Scientific EffectSignal transformation:

Implementation Method 3

followed by machine learning techniques such as support vector machines (SVM), auto-encoders, or convolutional neural networks (CNN) to analyze the status of annuli

Methodology Applied
Scientific EffectMachine learning pattern recognition:

Data Source

PatentEP3701124B1Methods of analyzing cement integrity in annuli of a multiple-cased well using machine learning
Publication Date: 2025.07.09 SERVICES PETROLIERS SCHLUMBERGER SA
  • EP3701124B1 patent drawingFigure 1
  • EP3701124B1 patent drawingFigure 1a
  • EP3701124B1 patent drawingFigure 1b

AI summary

A sonic tool is activated in a well having multiple casings and annuli surrounding the casing. Detected data is preprocessed using slowness time coherence (STC) processing to obtain STC data. The STC data is provided to a machine learning module which has been trained on labeled STC data. The machine learning module provides an answer product regarding the states of the borehole annuli which may be used to make decision regarding remedial action with respect to the borehole casings. The machine learning module may implement a convolutional neural network (CNN), a support vector machine (SVM), or an auto-encoder.