Flow regime classification in air-magnetic fluid two-phase flow

Area of Science:

• Fluid dynamics
• Two-phase flow analysis
• Electromagnetism

Background:

• Classifying flow regimes in two-phase flows is crucial for industrial processes.
• Traditional methods often require intrusive measurements, complicating analysis.
• Magnetic fluid applications necessitate understanding complex flow patterns.

Purpose of the Study:

• To develop and validate a non-contact technique for classifying air-magnetic fluid two-phase flow regimes.
• To utilize electromagnetic induction for signal acquisition and artificial neural networks for classification.
• To compare the accuracy of different artificial neural network models for flow regime identification.

Main Methods:

• Experimental setup for vertical upward air-magnetic fluid two-phase flow.
• Non-contact signal acquisition using electromagnetic induction to measure electromotive force.
• Signal processing via wavelet transforms.
• Supervised training of artificial neural networks (ANNs) with radial basis functions.
• Validation using a parallel visualization experiment with a glycerin solution.

Main Results:

• Time-series electromotive force signals were successfully obtained.
• Wavelet transforms effectively preprocessed the signal data.
• Artificial neural networks, particularly those with radial basis functions, demonstrated high accuracy in classifying flow regimes.
• Classification errors were minimized, validating the proposed technique.

Conclusions:

• The proposed non-contact method using electromagnetic induction and ANNs is effective for classifying air-magnetic fluid two-phase flow regimes.
• Radial basis function ANNs are optimal for this classification task.
• The technique offers a reliable and accurate alternative to intrusive measurement methods.