Study On The Feature Extraction Of Gas/liquid Two-Phase Flow Patterns And Information Fusion

In this study, the Lyapunov exponent, symbolic time series analysis method, D-S evidence fusion and their application in two-phase flow pattern identification are studied in stress. Firstly, the Lyapunov exponent is applied for the measurement fluctuation signal analysis of two-phase flow and the problem of using the value to characterize the flow patterns is also discussed. Then 80 groups'differential pressure fluctuation signals of gas/liquid two-phase flow have been gathered and the mutual information & method of false nearest neighbors are used to figure out the time delay and correlation dimension. Finally the small data algorithm is used to compute the largest Lyapunov exponent under all flow conditions.The flow pattern identification is an important aspect of the two-phase flow measurement. So in this study symbolic time series analysis method is used to deal with the gas/liquid two-phase flow differential pressure fluctuation signals which have been gathered in dynamic experiments. The results show that the characteristic quantities of different flow patterns have great distinctness, which reflects the rule of the flow character in two-phase flow. Hence, corresponding characteristic quantities can be used as the effective indicator to characterize the two-phase flow patterns.Finnaly D-S fusion method is used to treat with the signals. Multisource and heterogeneous evidence fusion is carried through by the association of BP netural network model and two-level D-S evidence reasoning model. So the method is put forward that the reliability of BP netural network classification result is used for the input proof probability of D-S evidence reasoning. The application results show that the the method is valid and provide a usefull investigation for the flow pattern identification of gas/liquid two-phase flow.