Study On Segmentation And Regional Feature Extraction Methods Of Differential Pressure Signal Of Gas-Liquid Two-Phase Flow

Gas-liquid two-phase flow is of great significance for many fields such as industrial production and pipeline transportation.However,due to the limitations of the flow mechanism and the complex interaction of the two phases,the characteristics of gas-liquid two-phase flow are not clear.The main content and innovations of this dissertation are as follows:(1)In this dissertation,the differential pressure signal of gas-liquid two-phase flow is divided into three regions by clustering method.The flow characteristics of the two-phase flow are studied for each region.Three cluster centers,dividing points of high,medium and low intervals are obtained.Original differential pressure signal is divided into three differential pressure regions.The two-phase flow characteristics are further analyzed in the three different regions.(2)The complex flow pattern of gas-liquid mixture in the pipeline is composed of many micro-units,such as bubbles,gas plugs,gas slugs,liquid plugs and liquid film.The differential pressure data is qualitatively related to the micro-units mentioned above.However,the partition of differential pressure data results in the loss of differential pressure data.In order to analyze the correlation between the statistical characteristics of differential pressure distribution and flow pattern and void fraction,data is filled into three regions using four different interpolation methods in this dissertation.It is found that the group mean interpolation method is simple and easy to implement,which effectively improves the estimation of the data variation degree,but the smoothness and accuracy of the interpolation node are not as good as the cubic spline interpolation method.The linear interpolation method is simple,the calculation speed is fast,and the software is easy to implement.However,when a nonlinear relationship exists in data,the calculation accuracy will not be guaranteed,and the overall smoothness is not enough.The calculation accuracy and speed of Newton interpolation is not high,and the error is larger.The cubic spline interpolation is far from the precision,which is easy to cause the dragon phenomenon.The cubic spline interpolation curve has good smoothness and good linear approximation ability.The interpolation results show that the cubic spline interpolation method has better effect in time and calculation accuracy,which is better than the other three methods.(3)The time-domain statistical characteristics of differential pressure signal in three regions are calculated.The relationships between the statistical characteristics,flow pattern and void fraction are analyzed.Kurtosis and skewness indices of the high differential pressure region can be used as a basis for flow pattern classification.The turbulence characteristic quantity,information entropy,and herd effect value,describing the degree of chaos of the system can be used as a basis for flow pattern classification in the middle differential pressure region.The fluctuation coefficient of low differential pressure region can be used as a basis for flow pattern classification.(4)The frequency-domain characteristics of differential pressure signal in three regions are analyzed.Amplitude-frequency maps,power spectral density maps and time-frequency map show that the frequency-domain characteristics of low differential pressure region can be used to identify flow patterns.