Flow regime information can be used to enhance measurement accuracy of flowmeters. Void fraction measurement and regime identification of two-phase flows including, liquid and gas phases are crucial issues in oil and gas industries. In this study, three different regimes including annular, stratified and homogeneous in the range of 5%–90% void fractions, were simulated by Monte Carlo N-Particle (MCNP) Code. In simulated structure, a Cesium 137 source and only one NaI detector were used to record received transmitted photons. Fast Fourier Transform (FFT) was applied to the registered signals of the detector in order to analyze in the frequency domain. Several features of signals in the frequency domain were extracted. These features were the average value of fast Fourier transform, the amplitude of dominant frequency, variance, Kurtosis and RMS (root mean square). Different combinations of these features were investigated in order to find the best features with the best separation ability for using as the inputs of Artificial Neural Network (ANNs). Two different Multi-Layer Perceptron (MLP) neural networks were used to recognize flow regimes and predict the void fraction. In regime identification procedure, all of the three mentioned regimes were recognized correctly and in the volume fractions prediction procedure, the void fraction was also estimated with a Mean Relative Error (MRE) percentage of less than 0.5%. In all of the previous studies, at least two detectors were used. Using the proposed method in this paper, number of detectors was reduced to one.
