Study On Characteristic Of Gas-liquid Two Phase Flow Measured Base On Dynamic Differential Pressure Signal

Gas-liquid two phase flow phenomenon exist widely in nature and allaspects of people’s production and lives. Therefore, study on dynamic characteristicsdeeply of gas-liquid two phase flow has important significance. As one of the mostimportant dynamic characteristics of two phase flow, the flow pattern is the mainresearch point and object in two phase flow field. The V-cone flowmeter as thethrottling device in experiments to acquire dynamic differential pressure signal.Local mean decomposition(LMD) algorithm is used to analysis differential pressuresignal and then extract the characteristic values of signal. By support vectormachine(SVM) training the characteristic values obtain classification model andrealize the flow pattern identification correctly. The specific studies andachievements are as follows:(1)Building the horizon pipeline experimental system of two phase flow, usingthe V-cone flow meter as the throttling device, collecting dynamic differentialpressure signal of two phase flow in different flow point by differential pressuretransmitter and data acquisition card. This paper mainly studies three typical flowpatterns, which is bubble flow, slug flow and plug flow. In experimental process wecollected298groups data.(2)Write a signal processing program to realize preprocessing of differentialpressure signal. The preprocessing include signal normalization, reject singular point,detrended, mathematical morphology filter(MMF) etc. After the signal preprocessingis more smooth, noise and power frequency interference is effectively suppressed.(3)The differential pressure signal of two phase flow can be separatedadaptively by LMD algorithm from high frequency band to low frequency band. Theproduct function(PF) is separated with a real physics meaning and preserve thecharacteristic of original signal, furthermore, has a lower end effects and modalaliasing.(4)Extract characteristic values of differential pressure signal, including energyentropy, PF energy, kurtosis coefficient and the standard deviation. Training the characteristic values by support vector machine(SVM), and then input the unknowntwo phase flow signal, so that it can be correctly identified. This paper uses30groups of samples for training,268groups of differential pressure signal foridentification, and the recognition is up to97.01%.(5)Write the system software programs based on LabVIEW, realize LMD algorithm,feature values extraction, storage and flow pattern recognition.