Analysis and identification of gas-liquid two-phase flow pattern based on multi-scale power spectral entropy and pseudo-image encoding

Abstract

Gas-liquid two-phase flow is closely related to the production and transportation of energy industries. A flow pattern analysis and identification method based on multi-scale power spectral entropy (MPSE) with pseudo-image encoding (PIE) is proposed. The resistance sensor array is used to collect gas-liquid two-phase flow information from the vertical upward pipeline, and the data is dimensionally reduced to simplify analysis. The MPSE is derived from the power spectrum of reduced-dimensional sequences to characterize the frequency domain complexity of the flow system at different scales. The optimal scale for spectral entropy analysis is further calculated to investigate the transition mechanism of gas-liquid two-phase flow patterns. In addition, the PIE algorithm is used to map the power spectrum into two-dimensional grayscale images to visualize the differences in frequency domain features of flow patterns. Combined with the deep learning model ResNeXt, the flow patterns are classified and compared with the traditional identification scheme. Results show that the proposed method can reveal the dynamic evolution behavior of gas-liquid two-phase flow from the frequency domain and achieve accurate identification of the flow patterns.