Title :
Flow Pattern Identification of Gas/Water Two Phase Flow Based on SVM
Author :
Zhao, Xin ; Jin, Ningde ; Zhang, Junxia
Author_Institution :
Sch. of Electr. Eng. & Autom., Tianjin Univ.
Abstract :
Based on the conductance fluctuating signals measured from gas/water two-phase flow in vertical upward pipe, the 10 feature quantities, which reflect flow characteristics of gas/water two-phase flow, were extracted. The features in frequency domain were derived by using linear prediction method of speech signal processing and the features in time domain were derived by time series statistical analysis. The extracted features were combined with the experimental observed flow pattern information to establish the data set, and the support vector machine was used to identify the flow patterns of gas/water two phase flows. Finally the classification validity reaches 95.39% for the three type flow patterns of bubble, slug and churn. We conclude that the support vector machine is an effective method to identify the flow patterns of gas/water two phase flows at the condition of small sample data
Keywords :
computational fluid dynamics; pattern classification; pipe flow; support vector machines; two-phase flow; conductance fluctuating signal; flow pattern identification; flow pattern information; frequency domain feature extraction; gas/water flow characteristics; gas/water two-phase flow; linear prediction method; support vector machine; time domain feature extraction; time series statistical analysis; Data mining; Feature extraction; Fluid flow measurement; Frequency domain analysis; Prediction methods; Speech analysis; Speech processing; Support vector machine classification; Support vector machines; Water; SVM; feature selection; flow pattern identification; two-phase flow;
Conference_Titel :
Intelligent Control and Automation, 2006. WCICA 2006. The Sixth World Congress on
Conference_Location :
Dalian
Print_ISBN :
1-4244-0332-4
DOI :
10.1109/WCICA.2006.1714158
