摘要：针对两相流体流动特性复杂、流型识别准确率低等问题，提出一种能够提高两相流流型识别率的方法。首先采用小波包分析对ERT 系统测量的压差波动信号进行特征提取；然后通过计算类间不可分离程度为每个节点选取最易分的两类构造DDAG 支持向量机多类分模型；最后将特征数据输入分类模型进行流型识别。通过实验对比，四种流型识别的准确率要明显高于其它常用方法的流型识别。结果表明，小波包分析和DDAG 支持向量机多类分类算法较大提高了油/水两相流流型识别的精度，是一种有效的流型识别方法。
中图分类号： TP391.4 文献标识码：A 文章编号：1672-9870（2015）04-0159-04
Application of DDAG Support Vector Machine in Flow Regime
Identification for Electrical Resistance Tomography System
（The City College of Jilin Jianzhu University，Changchun 130011）
Abstract：According to the fact that two-phase fluid has complex flow characteristic，and the accuracy of flow regime is low. In this paper，a method of improving recognizing rate of flow regime is presented. Firstly，wavelet packet analysis is adopted to extract the feature of the differential pressure fluctuation signal which is measured by electrical resistance tomography system，then the improved DDAGSVM muliticlass model is structured according to computing the inter- class separability which can distinguish two class easily for each mode. Finally the extracted feature data is taken as input information of the multi-class support vector machine of improved DDAG，so the four kinds of two-phase flow regime can be identified. Through experiment comparing，the accuracy rate of flow regime identification in this paper is higher than other method. Results show that the precision of two-phase flow regime identification is improved by the method of the wavelet packet analysis and DDAG support vector machine. It is an effective method of regime identification.
Key words：electrical resistance tomography；flow regime identification；wavelet packet；decision directed acyclic graph support vector machine