Flow pattern identification of gas-liquid flow based on the hybrid model of multi-scale information entropy feature and LS-SVM

Based on the characteristic that the Empirical Mode Decompostion (EMD) can decompose signal adaptively, a flow pattern identification method based on EMD multi-scale information entropy was put forward. Firstly, the acquired pressure-difference fluctuation signals are decomposed through EMD, and the decomposed signals within different frequency bands are obtained adaptively. Secondly, the multi-scale information signal entropy eigenvectors of flow pattern are abstracted. Finally, those eigenvectors are fed into the established hybrid model of EMD and LS-SVM for flow pattern identification and thus the flow pattern intelligent identification is realized. The experimental results show that this method can precisely identify the flow patterns of bubble flow, plug flow, and churn flow, respectively.