Nowadays, one of the most important research topics in the fields of oil, gas and petroleum industry is recognizing the type of flow regime and measuring the volume fraction of each component in multiphase flows. One of the methods for measuring multiphase flows is use of gamma ray attenuation technique. Compared to other measuring devices, this system has a lot of advantages such as: fast measuring, low cost and high portability and accuracy. Continuing or stopping drilling operations depends on the volumetric percentage of each component of the multiphase flows. The flow regime or flow pattern in a multiphase flows depends on a number of factors including: the density of each phase, difference in viscosity and velocity between phases. Sufficient information on components of multiphase flows will facilitate the separation operation. Due to the high cost of operations in the oil and gas industry, having accurate information on the volume fraction of each component and type of flow regime is very useful in profitability. In current study, to determine the type of flow regimes and predicting the volumetric fraction of each component of liquid-gas two-phase flows, three commonly flow regimes, namely: annular, stratified and homogeneous are simulated in void fraction of 5% to 90% by the Monte Carlo N Particle X-version (MCNP-X) code. In this thesis two laboratory structures are simulated. The first metering system consists of one 137Cs source, a Pyrex pipe with inner diameter of 95mm and thickness of 2.5mm and two sodium iodide detector. All components of the second structure are the same as the first, except that only one detector is used to simplify the system and reduce costs. Registered signal from detectors have a high frequency noises so, to tackle this destructive factor, the Savitzky-Golay filter was applied. After that, several features in time domain were extracted. Then the best separator features were chosen as the inputs of neural networks. Two type of Arti
