Fast recognition of global flow regime in pipeline-riser system by spatial correlation of differential pressures

Fast recognition of unstable and harmful flow regimes in pipeline-risers can promote a higher level of flow assurance in offshore petroleum exploitation. However, challenges are encountered in extracting distinguishable characteristics from a short sample which usually covers much less than a fluctuation period. This article attempts to set up a fast recognition method through the interrelation of differential pressures representing phase distribution along different sections, by which the regime category can be highly decoupled from time, and eventually sample length can be reduced remarkably with less complicated signal and data processing. Based on a number of experiments under different pressure conditions, we established a method that combined conditional judgment with LS-SVM whose inputs were dimensionless means and amplitudes of differential pressure signals, and four global regimes were conveniently recognized. The results proved the feasibility of our method, and time consumption was significantly reduced compared with that in literature. The shortest limit of sample length for effective recognition was discussed, and the sensitivity of trained model to test sample length was also analyzed through the distribution of input features.