【摘要】 根据小波包变换能将信号按任意时频分辨率分解到不同频段的特性,提出一种基于小波包多尺度信息熵的流型识别方法。该方法首先对采集到的压差波动信号进行4层小波包分解,在通频范围内得到分布在不同频段内的分解信号,进而建立流型的多尺度信息熵特征向量。并以此特征向量作流型样本对RBF神经网络进行训练,实现流型的智能化识别。试验结果表明,训练成功的RBF网络能很好地识别水平管内的4种流型,为流型识别开辟了一条新的途径。更多还原

【Abstract】 Based on the characteristic that the wavelet packet transform can decompose signal to different frequency bands in any time frequency resolution,a flow regime identification method based on wavelet packet multiscale information entropy is proposed.The collected pressure-difference fluctuation signals are decomposed into four-layer wavelet packets,and the decomposed signals in various frequency bands are obtained within the pass frequency band.Then the multi-scale information entropy eigenvectors of flow regime are established.The RBF neural network is trained using those eigenvectors as flow regime samples and the flow regime intelligent identification is realized.The test result shows after successful training the RBF neural network can identify four flow regimes of gas-water two-phase flow in horizontal pipe.This method introduces a new direction for flow regime identification.更多还原