Determination of Two-Phase Flow Void Fraction of R600a in a Horizontal Smooth Tube

Abstract

In this paper, the void fraction of alternative refrigerant R600a flowing inside horizontal tube is determined by means of an experimental technique, well known correlations in the literature and a generalized neural network analysis. The horizontal tube is made from smooth glass tubing of 4 mm inner diameter. The test runs are done at average saturated condensing temperatures between 30 and 40 °C while the average qualities and the mass fluxes are between 0.45–0.91 and 68.5–138.1 kg m-2s-1 respectively. The flow regime determination inside the tube is performed by means of sight glasses placed at the inlet and outlet sections of the test section, used for in-tube condensation tests, virtually. An image processing technique, performed by means of a high speed camera, is used to determine the void fractions of stratified and annular condensing flow of R600a experimentally. The void fractions are determined using relevant measured data together with 11 different void fraction models and correlations reported in the open literature analytically. Artificial neural network (ANN) analysis is developed to determine the void fractions numerically. For this aim, mass flow rate, average vapor quality, saturation temperature, liquid and vapor densities, liquid and vapor dynamic viscosities and surface tension are selected as the input parameters, while the void fraction is selected as the output. Three-layer network is used for predicting the void fraction. The number of the neurons in the hidden layer was determined by a trial and error process evaluating the performance of the network and standard sensitivity analysis. The measured void fraction values are found to be in good agreement with those from ANN analysis and correlations in the literature. It is also seen that the trained network are more predictive on the determination of void fraction than most of the investigated correlations.