基于小波包多尺度信息熵和HMM的气液两相流流型识别方法

为了研究垂直上升管中气液两相流的流型,利用自制的多电导探针测量系统采集了四种典型流型的电导波动信息.由于气液两相流电导波动信号的非平稳特征,提出了一种基于小波包多尺度信息熵(Wavelet Packet Multi-scal2e Information Entropy)和隐马尔可夫模型(Hidden Markov Model,HMM)的流型识别方法.该方法首先对采集到的电导波动信号进行3层小波包分解,得到了8个不同频带的信号,提取各频带信号的小波包多尺度信息熵特征作为流型的特征向量,然后将其转换为观测序列输入到各种状态的隐马尔可夫模型进行训练并识别流型.结果表明:与BP神经网络相比,采用隐马尔可夫模型进行流型识别可以获得更高的识别率,表明该方法是有效和可行的.

A method for identifying gas-liquid two-phase flow patterns on the basis of wavelet packet multi-scale information entropy and HMM

For studying flow regimes of gas/liquid two-phase in a vertical upward pipe, the conductance fluctuation information of four typical flow regimes was collected by a measuring the system with self-made multiple conductivity probes. Owing to the nonstationarity of conductance fluctuation signals of gas-liquid two-phase flow, a kind of flow regime identification method based on wavelet packet Multi-scale Information Entropy and Hidden Markov Model(HMM)was put forward. First of all, the collected conductance fluctuation signals were decomposed into eight different frequency bands signals. Secondly, the wavelet packet multi-scale information entropy of different frequency bands signals were regarded as the input characteristic vectors of all states HMM which had been trained. In the end the regime identification of the gas-liquid two-phase flow could be performed. The study showed that the method that HMM was applied to identify the flow regime was superior to the one that BP neural network was used, and the results proved that the method was efficient and feasible.