Artificial neural network analysis for performance of parallel connected vortex tubes

Abstract

In this study, two counter-flow Ranque-Hilsch Vortex Tubes (RHVTs) were connected in parallel, and their performance was investigated experimentally, and the temperature difference (?T) values the vortex tube's performance indicator between hot and the cold fluid outlet were obtained. In experiments with oxygen and air, 3,4 and 5 orifice nozzles made of polyamide and brass are used. An artificial neural network (ANN) study was conducted to model the ?T values obtained for different fluids, nozzle materials and nozzle numbers, and generalizable modeling was obtained for two parallel vortex tube systems. Thermal conductivity and orifice number for nozzles, specific heat and density parameters for working fluids and RHVT inlet pressure (5 input) are used as input parameters. Data for ANN was separated as a training and test group and the trained model was tested with the test group. In regression analysis, R2 value was calculated as 99.8% for the education group and 99.6% for the test group. © 2020 Elsevier B.V., All rights reserved.