基于压力信号统计特征的管道式离心分离器内流型失稳识别OA

To address the shortcomings of existing methods in noise sensitivity and parameter dependency for judging flow pattern instability in vertical gas-liquid swirling flows,a flow pattern identification method is proposed that integrates kernel density estimation(KDE)and dimensionless parameters.First,the dynamic evolution of flow patterns was observed through a visualization experiment system.Based on an improved KDE method,time-domain statistical features of pressure signals were extracted.Dimensionless parameters such as standard deviation(σ),skewness coefficient(Sₖ),coefficient of variation(CV),and kurtosis(K)were introduced to establish a joint criterion for quantifying flow pattern differences.Finally,the criterion and transition boundary for unstable swirl-ing flow patterns were determined.Experiments show that the transition boundary follows an exponential relation-ship:jl=1.177 4 j0.63g(jg is the gas-phase superficial velocity;jl is the liquid-phase superficial velocity).This result provides a quantitative criterion for flow pattern instability warning.The proposed method can offer reliable technical support for online flow pattern monitoring,instability warning,and operational optimization of centrifugal separation devices.