Void fraction prediction in two-phase flows independent of the liquid phase density changes

Abstract

Gamma-ray densitometry is a frequently used non-invasive method to determine void fraction in two-phase gas liquid pipe flows. Performance of flow meters using gamma-ray attenuation depends strongly on the fluid properties. Variations of the fluid properties such as density in situations where temperature and pressure fluctuate would cause significant errors in determination of the void fraction in two-phase flows. A conventional solution overcoming such an obstacle is periodical recalibration which is a difficult task. This paper presents a method based on dual modality densitometry using Artificial Neural Network (ANN), which offers the advantage of measuring the void fraction independent of the liquid phase changes. An experimental setup was implemented to generate the required input data for training the network.ANNs were trained on the registered counts of the transmission and scattering detectors in different liquid phase densities and void fractions. Void fractions were predicted by ANNs with mean relative error of less than 0.45% in density variations range of 0.735 up to 0.98 gcm−3. Applying this method would improve the performance of two-phase flow meters and eliminates the necessity of periodical recalibration.