Flow Regime Identification in Large Diameter Pipe

Abstract

Air-water vertical two-phase flow experiments were performed in a 0.15 m diameter and 4.4 m long test section. Superficial liquid velocity was varied from 0.05 m/s to 2.0 m/s and superficial gas velocity was varied to obtain the area averaged void fraction range of 0.1 to 0.7. Exit pressure was close to the atmospheric pressure. In order to study the development of flow structure over the length of test section, area averaged void fraction was measured using impedance meters at four different measuring ports. Pressure drop was also measured between these ports. Since the temporal variation of void fraction signal obtained from the impedance meter and its distribution are characteristic of the flow regime, a Cumulative Probability Distribution Function (CPDF) of the void fraction signal was utilized for the identification of flow regime at each port. The CPDFs of the impedance probe void fraction signal were supplied as an input to the Kohonen Self Organized neural network or the Self Organized Map (SOM) for the identification of the patterns by employing self organized neural network technique. The three flow regimes identified by the neural network are subjectively named as bubbly flow, cap-bubbly flow and cap-turbulent flow.