基于独立分量分析和支持向量机的气液两相流流型识别  

A New Approach to Identify the Two-phase Flow Patterns by Combined ICA with SVM

作　　者：周云龙[1] 顾杨杨[1] 

机构地区：[1]东北电力大学能源与动力工程学院,吉林吉林132012

出　　处：《东北电力大学学报》2012年第1期1-6,共6页Journal of Northeast Electric Power University

基　　金：国家自然科学基金项目(50976018);吉林省自然科学基金项目(20101562)

摘　　要：为了准确的识别气液两相流流型,提出了将独立分量分析与支持向量机相结合的流型识别方法。首先应用独立分量分析对流型的压差波动信号进行特征参数提取,然后将参数通过支持向量机进行分类处理,从而实现了对流型的识别。经试验证明:该方法可以准确的识别出水平管内的四种流型,整体识别率可以达到96.7%,为流型的在线识别提供了一种新的技术依据。In order to identify the two-phase flow patterns accurately,propose a new approach which is combined ICA with SVM.First,used independent component analysis extract the characteristic parameters from differential pressure fluctuation of flow pattern signals,then put the parameters into SVM in order to processing classification,so achieved the purpose of identifying flow patterns.Experimental proof that：this method can identify four kinds of flow patterns in horizontal pipe accurately,the whole identification accuracy was 96.7%,this method also provides a new technical means for on-line flow identification.

关 键 词：独立分量分析 盲源分离 特征参数 支持向量机 

分 类 号：O359.1[理学—流体力学]