An artificial intelligence based improved classification of two-phase flow patterns with feature extracted from...

Area of Science:

• Engineering
• Fluid Dynamics
• Computational Science

Background:

• Accurate flow pattern recognition is crucial for process design and simulation in two-phase flow systems.
• Automating flow pattern interpretation using visual image processing offers significant advantages.

Purpose of the Study:

• To improve the classification accuracy of gas/liquid two-phase flow patterns.
• To evaluate the effectiveness of fuzzy logic, Support Vector Machine (SVM), and SVM with Principal Component Analysis (PCA) for flow pattern identification.

Main Methods:

• Recorded videos of six flow patterns (annular, bubble, churn, plug, slug, stratified) and converted them to 2D images.
• Extracted textural and shape features from images using image processing techniques.
• Applied fuzzy logic, SVM, and SVM with PCA as classification schemes.

Main Results:

• Support Vector Machine (SVM) combined with Principal Component Analysis (PCA) demonstrated superior classification accuracy.
• The SVM with PCA approach proved to be more computationally efficient than fuzzy logic and standard SVM.
• The developed method shows promise for industrial applications in oil and gas and other gas-liquid two-phase flow scenarios.

Conclusions:

• SVM with PCA offers an effective and efficient solution for automated gas/liquid two-phase flow pattern recognition.
• The study provides a valuable tool for enhancing process design and simulation in relevant industries.
• This approach addresses the need for accurate and computationally feasible flow pattern identification.