针对起伏振动条件下气液两相流压差信号过于复杂难以识别的问题，提出一种基于改进的自适应噪声完备集合经验模态分解(ICEEMDAN)与支持向量机（SVM）相结合的流型识别方法。采用ICEEMDAN对小波去噪后的压差信号进行模式分解，通过求取的各本征模态函数（IMF）与原始信号进行斯皮尔曼相关系数计算，选取相关系数较大的IMF分量进行希尔伯特变换，对变换后各IMF分量的瞬时幅度进行能量熵、奇异谱熵、功率谱熵的计算，构成特征向量，带入到支持向量机中进行流型识别。结果表明：该方法能够有效识别起伏振动状态下的泡状流、弹状流、搅混流、环状流，识别准确率可达95%。

In order to solve the problem that the differential pressure signals of gasliquid twophase flow under fluctuating vibration conditions were too complicated and difficult to identify, a flow pattern identification method based on the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and support vector machine (SVM) was proposed. The ICEEMDAN was used to decompose the pressure difference signal after wavelet denoising. The Spearman correlation coefficient was calculated by the obtained each intrinsic mode functions (IMF) and the original signal. The IMF component with large correlation coefficient was selected for Hilbert transformation. The instantaneous amplitude of each transformed IMF component was calculated by energy entropy, singular spectrum entropy and power spectrum entropy, and then the feature vector was formed and brought into the support vector machine for flow pattern identification. The results show that this method can effectively identify bubble flow, slug flow, churn flow and annular flow in the state of fluctuating vibration, and the accuracy of identification can reach 95%.

O359.1

国家自然科学基金（51776033）