Flow Pattern Identification Based on EMD and LS-SVM for Gas–Liquid Two-Phase Flow in a Minichannel

Based on empirical mode decomposition (EMD) and least squares support vector machine (LS-SVM), a new method is proposed to identify the flow pattern of gas-liquid two-phase flow in a minichannel. Four flow patterns are observed in three pipes with inner diameters of 4.0, 3.1, and 1.8 mm. For each flow pattern, the capacitance signals are obtained by a two-electrode capacitance sensor. The EMD method is applied to the capacitance signal to obtain intrinsic mode functions (IMFs) with different characteristic time scales. For each IMF, the autoregression (AR) model is built to extract multiscale features. Combining the extracted features with the energy feature of each IMF, the flow patterns are identified by the multiclassification LS-SVM classifier. The experimental results indicate that the presented method is effective for flow pattern identification and has identification rates higher than 91%.