Flow regime identification of swirling gas-liquid flow with image processing technique and neural networks

Abstract

Swirling flow is one of the commonly-recognized techniques to control working processes in various engineering fields. A fundamental understanding of the swirling flow pattern is significantly important for proper design, operation and optimization of swirling flow systems. Although a large amount of work has been done on conventional non-swirling two-phase flow, investigations on swirling flow regime classification are still scarce. In this paper, a visualization experiment was carried out to study the gas-liquid flow in a vertical pipe containing a swirler with four helical vanes. The typical flow regimes in swirling gas-liquid flow were first classified and defined by visual observation. Subsequently, the liquid holdup was measured with the proposed image processing technique and statistically analyzed through PDF and CPDF. Owing to its integral and stable feature, CPDF was then utilized along with a self-organizing neural network (SONN) to identify the swirling flow regimes objectively. Finally, a swirling flow regime map was proposed and compared with the previous non-swirling and swirling maps available in the publications. Based on this study, the influence of the centrifugal force on phase distribution of gas-liquid flow was clarified and the general relationship between swirling gas-liquid flow and non-swirling gas-liquid flow was concluded.