基于HMM和小波包分解的气液两相流流型识别  

Identifying Gas-liquid Two-phase Flow Patterns Based on HMM and Wavelet Packet Decomposition

作　　者：周云龙[1] 张学清[2] 李峰[1] 

机构地区：[1]东北电力大学能源与机械工程学院,吉林吉林132012 [2]东北电力大学自动化工程学院,吉林吉林132012

出　　处：《自动化仪表》2009年第8期42-46,共5页Process Automation Instrumentation

基　　金：吉林省科技发展计划基金资助项目(编号:20040513)

摘　　要：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4种典型流型的电导波动信息,提出了基于HMM和小波包分解的气液两相流流型识别方法。首先应用小波包分解对电导波动信号进行小波包能量特征参数的提取,然后将小波能量参数作为观测序列输入到隐马尔科夫模型(HMM),从而实现对流型的识别。研究结果表明,该方法能够准确地识别出4种流型,识别效果良好,为流型的在线识别提供了一种有效方法。In order to research the patterns of gas-liquid two-phase flow in vertical upward pipelines, by adopting self-made measuring system with multiple conductance probes, the conductance fluctuation information of four types of typical flow patterns is collected, and the identification method based on HMM and wavelet packet decomposition is proposed. First, conductance fluctuation signals are extracted for wavelet energy feature parameters by using wavelet packet decomposition; then these parameters are put into HMM as observation sequence, thus the identification of flow patterns is implemented. The result of research shows that by using the method, four patterns can he identified properly with well effect and it provides an effective measure for online flow patterns identification.

关 键 词：小波包分解 电导探针 两相流 流型识别算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]