Camera-Based Machine Learning for Fluid Phase Separation Chemistry Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for assessing the effectiveness of phase separation chemicals in fluid samples from subsurface formations suffer from poor repeatability and are subject to human bias, leading to inconsistent quality information and time-consuming processes.
Innovation Solution
Utilizing a learning machine, such as a convolutional neural network (CNN), to analyze media content of fluid samples captured by cameras, determining properties like water quality, interface quality, and water drop volume to assess the effectiveness of phase separation chemicals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional human-based visual examination methods are used to assess fluid sample properties, then expert knowledge and experience can be utilized, but the process becomes time-consuming and subject to human bias leading to poor repeatability
Solution Approach 1:
The patent replaces the mechanical human visual examination system with an automated optical system comprising a camera, light source, and image processing algorithms. The system captures images of the fluid sample and uses computer vision techniques to automatically determine water drop volume, water quality, and interface quality, eliminating human bias and time-consuming manual assessment while maintaining or improving measurement precision
Solution Approach 2:
The system enables self-service by allowing the fluid sample to be automatically analyzed without requiring expert intervention. The automated image processing and machine learning algorithms independently assess the sample properties, making the evaluation process independent of human expertise while improving consistency and reducing time loss
2Reliability
If human experts conduct visual examination of fluid samples, then quality information can be obtained, but the process suffers from subjectivity and poor repeatability
Solution Approach 1:
The patent replaces the unreliable human expert system with a reliable automated optical system. The camera-based imaging system combined with standardized image processing algorithms provides objective, repeatable measurements of fluid sample properties, eliminating subjectivity while the modular system design keeps complexity manageable
Solution Approach 2:
The system changes the assessment parameters from subjective human judgments to quantifiable image-based metrics. By converting visual properties into measurable parameters such as water drop volume, water quality, and interface quality through image analysis, the system achieves high repeatability while maintaining reasonable complexity through standardized measurement protocols
3Productivity
If manual visual assessment methods are used, then simple equipment can be utilized, but the evaluation process is time-consuming and less frequent
Solution Approach 1:
The patent replaces manual visual assessment with an automated optical analysis system that can process multiple images and evaluate numerous fluid samples rapidly. The computer-based image processing and machine learning algorithms enable high-throughput analysis, significantly increasing evaluation frequency while the modular architecture keeps system complexity manageable
Solution Approach 2:
The system enables continuous evaluation by automatically capturing and analyzing fluid sample images without interruption. The automated workflow allows for continuous monitoring and assessment of multiple samples, maximizing productivity while the streamlined processing keeps device complexity at acceptable levels
Data Source
AI summary
A method comprises obtaining a fluid sample produced from a subsurface formation. The method comprises loading the fluid sample into a device configured with one or more cameras. The method comprises obtaining, via the one or more cameras, media content of the fluid sample. The method comprises determining, via a learning machine, one or more sample properties of the fluid sample based on the media content.