Intelligent identification of steam jet condensation regime in water pipe flow system by wavelet multiresolution analysis of pressure oscillation and artificial neural network

• Published in: Applied thermal engineering 2019-01, Vol.147, p.1047-1058
• Main Authors: Xu, Qiang, Ye, Shuyan, Liu, Weizhi, Chen, Yanshuang, Chen, Qiyu, Guo, Liejin
• Format: Article
• Language: English
• Subjects: Artificial neural network; Artificial neural networks; Condensation; Condensation regime recognition; Multiphase flow; Multiresolution analysis; Neural networks; Pipe flow; Pressure oscillation; Pressure oscillations; Signal processing; Steam jets; Vapor jets; Water pipelines; Water pipes; Wavelet analysis; Wavelet multiresolution analysis
• ISSN: 1359-4311; 1873-5606
• DOI: 10.1016/j.applthermaleng.2018.11.005

•Comprehensive database of jet regimes associated with pressure signal is established.•Wavelet analysis is used to extract features of pressure signal for classification.•Jet regimes exhibits clear interrelation with wavelet features of pressure signal.•Artificial neural network is applied to the clusters for constitution of classifiers.•Satisfactory identification rate of jet regimes by pressure signal is obtained. On-line recognition of condensation regime of vapor jet in pipe flow systems is a promising approach for flow assurance and intellectualization of industrial processes. However, the selection of distinguishable characteristics from pressure signals associated strongly with various condensation regimes is essential and challenging for satisfactory recognition purpose. Accordingly, an artificial neural network technique using wavelet multiresolution analysis of pressure oscillation signals for objective identification of jet condensation regimes is presented in this paper. The recognition procedure was carried out in two major steps. Statistical features of wavelet multiresolution analysis of pressure signals, i.e., mean of absolute and percentage of energy of each wavelet scale, were chose first. And then artificial neural network was adopted to construct classifiers for forecasting the condensation regimes automatically. The recognition results illustrated that the proposed method is feasible and effective for identifying vapor jet condensation regime in pipe flow system. Furthermore, it is suggested that statistical features of mean of absolute and percentage of energy at least four or more particular wavelet scales, and also sample length longer than 1.5 s could guarantee a satisfactory recognition rate above 90%.