In this work, the Convolutional Neural Network VGG-16 is applied for the analysis of histogram images of signals obtained for liquid-gas flow by using gamma-ray absorption. The experiments were carried out in the laboratory hydraulic installation with horizontal pipeline fitted with a radiometric measurement system. The radiometric measurement set used in the investigation consists of a linear Am-241 radiation source with an energy of 59.5 keV and NaI(Tl) scintillation detector. In this work four types of water-air flow regimes as plug, slug, bubble, and transitional plug–bubble were studied. MATLAB software was used to analyze the measurement signal obtained from the detector. It was found that the VGG-16 network correctly recognizes the flow regime in more than 90% of the cases.

Knowledge of the two-phase flow structure is essential for the proper conduct of industrial processes. Description of liquid-gas flow regimes is possible by using of data analysis in time, frequency, or state-space domain. In this research studies, the autocorrelation function is applied for analysis of signals obtained for liquid-gas flow by use gamma-ray absorption. The experiments were carried out on the laboratory hydraulic installation fitted with Am-241 radioactive source and scintillation probe with NaI(Tl) crystal. Four types of flow regimes as plug, slug, bubble, and transitional plug - bubble were studied in this work. It was found that the selected amplitudes of the normalized autocorrelation function are helpful to recognize the flow regime.

In this paper, the use of the gamma-absorption method applied in the investigation of the two-phase liquid-gas flow in the pipeline is described. An example of its application to the air transported by water in a horizontal pipeline is evaluated. In the measurements, Am-241 radioactive sources and probes with Nal (Tl) scintillation crystals have been used. The signals from the radiometric set were used to determine the velocity of the dispersed phase, void fraction and for identification of flow regime. In the latter case, artificial neural networks were applied. Results of the selected experiments carried out on a laboratory stand are presented.