A NOVEL METHOD FOR GAS-LIQUID FLOW PATTERN RECOGNITION BASED ON PDF AND SVM

Currently, most of the traditional flow pattern recognition methods are based on the features of differential pressure signal, and the verification data is majorly derived from the horizontal or almost horizontal pipes. This paper proposes a new method that combines the probability density function (PDF) of the liquid holdup signal and the support vector machine (SVM). Results demonstrate the capability of the proposed algorithm to effectively recognize the stratified flow, bubble flow, slug flow, and severe slug flow and eliminate subjective factors.