Two-Phase Air–Water Slug Flow Measurement in Horizontal Pipe Using Conductance Probes and Neural Network

This paper presents a method to obtain gas and liquid flow rates of two-phase air-water slug flow in a horizontal pipe through conductance probes and neural network. Contrary to statistical features commonly used in other works, five characteristic parameters of the mechanistic slug flow model are extracted from conductance signals, i.e., translational velocity, slug holdup, film holdup, slug length, and film length, which are used as the neural network inputs. The translational velocity is obtained through cross correlation of signals from the two ring-type conductance probes that are placed apart at a fixed distance. A feedforward neural network is adopted to correlate the characteristic parameters of slug flow and the gas and liquid flow rates and further used as a prediction tool. The experimental results show that the neural network method is able to learn the implicit correlations between the characteristic parameters of slug flow and the corresponding gas and liquid flow rates. It provides a performance for measurement of gas and liquid flow rates in slug flow regime within ±10% of full scale.