Identification of flow regime in vertical upward air–water pipe flow using differential pressure signals and elastic maps

Abstract

A new method for the identification of flow regime in vertical upward gas–liquid flows has been proposed. It consists of an application of the elastic maps algorithm, which is a machine learning method, to the probability density function of differential pressure measurements in pipes. The proposed method was found to be insensitive to axial location of the measurements, pipe diameter within the range 13.9–49.2mm and absolute pressure within the range 100–240kPa; it is therefore amenable to relatively simple calibration in a calibration rig. Compared to three previously suggested machine learning algorithms, the present one had a superior performance in identifying the gas–liquid flow regime.