Noninvasive Flow Regime Identification for Wet Gas Flow Based on Flow-induced Vibration  

作　　者：华陈权 王昌明 耿艳峰 石天明 

机构地区：[1]College of Information & Control Engineering, China University of Petroleum, Dongying 257061, China [2]College of Mechanical Engineering, Nanjing University of Science and Technology, Nanjing 210094, China

出　　处：《Chinese Journal of Chemical Engineering》2010年第5期795-803,共9页中国化学工程学报（英文版）

基　　金：Supported by the National Natural Science Foundation of China （60672003）

摘　　要：A novel noninvasive approach, based on flow-induced vibration, to the online flow regime identification for wet gas flow in a horizontal pipeline is proposed. Research into the flow-induced vibration response for the wet gas flow was conducted under the conditions of pipe diameter 50 mm, pressure from 0.25 MPa to 0.35 MPa, Lockhart-Martinelli parameter from 0.02 to 0.6, and gas Froude Number from 0.5 to 2.7. The flow-induced vibration signals were measured by a transducer installed on outside wall of pipe, and then the normalized energy features from different frequency bands in the vibration signals were extracted through 4-scale wavelet package transform. A "binary tree" multi-class support vector machine(MCSVM) classifier, with the normalized feature vector as inputs, and Gaussian radial basis function as kernel function, was developed to identify the three typical flow regimes including stratified wavy flow, annular mist flow, and slug flow for wet gas flow. The results show that the method can identify effectively flow regimes and its identification accuracy is about 93.3%. Comparing with the other classifiers, the MCSVM classifier has higher accuracy, especially under the case of small samples. The noninvasive measurement approach has great application prospect in online flow regime identification.A novel noninvasive approach, based on flow-induced vibration, to the online flow regime identification for wet gas flow in a horizontal pipeline is proposed. Research into the flow-induced vibration response for the wet gas flow was conducted under the conditions of pipe diameter 50 mm, pressure from 0.25 MPa to 0.35 MPa, Lockhart-Martinelli parameter from 0.02 to 0.6, and gas Froude Number from 0.5 to 2.7. The flow-induced vibration signals were measured by a transducer installed on outside wall of pipe, and then the normalized energy features from different frequency bands in the vibration signals were extracted through 4-scale wavelet package transform. A ＂binary tree＂ multi-class support vector machine（MCSVM） classifier, with the normalized feature vector as inputs, and Gaussian radial basis function as kernel function, was developed to identify the three typical flow regimes including stratified wavy flow, annular mist flow, and slug flow for wet gas flow. The results show that the method can identify effectively flow regimes and its identification accuracy is about 93.3%. Comparing with the other classifiers, the MCSVM classifier has higher accuracy, especially under the case of small samples. The noninvasive measurement approach has great application prospect in online flow regime identification.

关 键 词：flow regime identification wet gas flow flow-induced vibration wavelet package transform support vector machine 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TV698.162[自动化与计算机技术—计算机科学与技术]