Neural network analysis of void fraction in air/water two-phase flows at elevated temperatures

Radial basis function neural networks have been used to predict cross-sectional and time-averaged void fraction at different temperatures. The data bank contains experimental measurements for a wide range of operational conditions in which upward two-phase air/water flows pass through a vertical pipe of 2.42 cm diameter. The independent parameters are in terms of dimensionless groups such as modified volumetric flow ratio, density difference ratio, and Weber number. A comparison between the experimental and predicted data reveals an overall average error of 3.6% for training and 5.8% for unseen data. In addition, the trend of both predicted results and experimental data are qualitatively consistent.