Development of a Real-Time Objective Gas-Liquid Flow Regime Identifier Using Kernel Methods

Area of Science:

• Fluid dynamics
• Chemical engineering
• Instrumentation and measurement

Background:

• Traditional flow regime identification in closed channels relies on subjective visual methods.
• Opaque pipes and unclear transitions at specific gas-liquid flow rates pose significant challenges.
• A need exists for objective, real-time flow regime identification tools.

Purpose of the Study:

• To develop a novel, real-time, objective flow regime identification tool for horizontal annular closed channels.
• To utilize conductance data and advanced kernel methods for accurate flow regime classification.
• To overcome limitations of subjective methods in opaque or complex flow conditions.

Main Methods:

• A flush-mounted conductance probe was used to collect voltage signals from a horizontal annular channel.
• Kernel principal components analysis (KPCA) and multiclass support vector machine (SVM) were employed to train the identifier.
• The system processed voltage signal probability density functions (PDFs) within a moving time window for real-time analysis.

Main Results:

• The objective identifier achieved over 90% accuracy compared to expert visual observations on static data.
• Virtual flow regime maps were generated, effectively visualizing conductance data in 2-D space for transition tracking.
• An efficient, real-time automatic flow regime identifier was successfully developed using only conductance data.

Conclusions:

• The developed tool offers an accurate and objective method for gas-liquid flow regime identification in closed channels.
• Real-time monitoring and visualization capabilities enhance understanding of flow dynamics and transitions.
• This approach provides a valuable, data-driven alternative to traditional subjective identification techniques.