Gas-liquid Two-phase Flow Pattern Identification Based on Variational Mode Decomposition and Deep Learning

The identification of gas-liquid two-phase flow pattern is crucial for fluid dynamics research and industrial applications. We propose a novel flow pattern identification framework, combining variational mode decomposition (VMD), mutual information (MI), and deep learning. Firstly, the four-channel conductance signals are decomposed into intrinsic mode functions (IMFs) by VMD to capture subtle flow characteristics. Then, the statistical dependency between each IMF is calculated by MI. Subsequently, these dependencies are organized into a feature matrix, which effectively represents the nonlinear interactions among different components of the signals. Finally, the constructed feature matrix is fed into four models for gas–liquid flow pattern identification, achieving an accuracy of 95.56% on a self-constructed dataset. These result demonstrate the effectiveness of the proposed approach in capturing key flow characteristics and achieving accurate flow pattern identification.