Calculation for critical gas velocity of liquid accumulation in inclined pipelines: A method based on physics-informed neural network

Liquid accumulation in gas pipelines will reduce transportation efficiency, increase corrosion rates, and induce severe slug flow. Calculation for critical gas velocity v_cg of liquid accumulation in inclined pipelines is important for the prevention of liquid accumulation. Due to the complexity of multiphase flow, the mechanism of liquid accumulation is still controversial. Many models have been proposed based on different liquid accumulation theories, but most of these models are complex and inaccurate. It is difficult to compare the calculation results of different theories in a unified standard. To simplify the calculation and improve the accuracy, a new physics-informed neural network (PINN) for calculating v_cg is proposed. PINN is trained only by the physical constraints of gas-liquid two-phase flow (GLF) and does not require any training data. In the same computational framework, PINN can calculate the v_cg corresponding to minimum pressure gradient (MPG), minimum gas-liquid interface shear stress (MIS), and zero liquid-wall shear stress (ZLS), respectively. In addition, the same two empirical equations are introduced for each calculation procedure, which ensures objectivity in the evaluation of different liquid accumulation theories. With 89 collected public experimental data, PINN is compared with 3 models based on different theories, and the changing law of v_cg are analyzed. The results show that PINN is applicable to a range of operating conditions with liquid superficial velocity from 0.001 to 0.100 m/s, pipe inclination from 2° to 20°, and pipe diameters from 50 to 200 mm. PINN are better than other models, and different theories have different sensitivities to each factor. This study provides a new computational method for the research of GLF and provides guidance for the prevention of liquid accumulation in gas pipelines.