Study on velocity profile of gas–liquid two-phase stratified flow in pipelines based on transfer component analysis-back propagation neural network

The measurement of cross-sectional velocity profile is a challenge in the field of two-phase flow. In this paper, the stereoscopic particle image velocimetry (SPIV) technique is employed to obtain the cross-sectional velocity profile of gas and liquid phase in stratified flow. Interface velocity profile is obtained through numerical simulation. By leveraging the concept of transfer learning, we propose to construct a transfer component analysis-back propagation network using stereo particle image velocimetry and numerical simulation and to predict the velocity profile of the gas–liquid interface in stratified flow. The research indicates that the cross-sectional velocity profile of the gas–liquid stratified flow is similar to the “mushroom” shape. The velocity profile of the gas–liquid interface changes from an M-type to the N-type, and the gas–liquid velocity slip affects the transformation process. With the increase in the gas-phase velocity, the distance between the two peaks of the M-type velocity profile increases and the gap between gas–liquid velocity peaks increases accordingly.