Neural networks approach for prediction of gas–liquid two-phase flow pattern based on frequency domain analysis of vortex flowmeter signals

The identification of flow pattern is a basic and important issue in multiphase systems. Because of the complexity of phase interaction in gas–liquid two-phase flow, it is difficult to discern its flow pattern objectively. In this paper, a three-layer, feed-forward neural network was designed, and adopted inputs all representing the characteristics of the power spectral density distributions of dynamic differential pressure fluctuations were obtained from a vortex flowmeter. The validity of the adopted inputs for the flow pattern identification was evaluated by a proposed effectiveness factor. Results show that the designed neural networks predict the flow patterns successfully comparing with the flow pattern by visual observation. These findings provide the possibility of using only a vortex flowmeter for the identification of gas–liquid two-phase flow patterns with the help of neural networks.