基于隐马尔科夫模型的气液两相流流型识别方法

A Method for Identifying Gas-liquid Two-phase Flow Patterns Based on HMM

提出一种基于隐马尔科夫模型(HMM)识别垂直上升管中气液两相流流型的方法。首先对采集到的电导波动信号进行数据处理,借助语音信号处理中的线性预测方法,提取出反应两相流流动特征的特征量。将提取的特征向量作为观测序列输入到已经训练完毕的各状态HMM中,从而实现对气液两相流流型的识别。结果表明,该软测量方法能够准确地识别出三种典型的流型,识别效果良好,为流型的识别提供了一种有效的软测量方法。

A method for identifying gas-liquid two-phase flow pattern in vertical pipe based on Hidden Markov Model （HMM） is put forwards. First, the collected conductance fluctuation signals are processed to extract flow characteristic parameter of gas-liquid two-phase with the aid of linear forecasting method in the speech signal processing. These collected parameters are put into the finished state of training HMM to realize the gas-liquid two phases flow pattern recognition. The results show three typical flow patterns can be identified properly by using the soft measurement method, and it can provide a kind of effective method of the soft measurement for flow patterns identification.