Modeling the Performance of Vortex Tube using Response Surface Methodology and Artificial Neural Networks

Abstract

Vortex tube gives hot and cold streams of gas taking pressurized gas as input. Though this device is mainly used for spot cooling purposes, hot temperature of outlet gas as a response variable, is also of concern for experimentation. The present study deals with modelling and analyzing the effect of five input controllable parameters, each considered with two levels (Max. and Min) viz., internal Diameter of the hot tube in mm, Dt, Length of the hot tube in mm, L, inlet Pressure of air, Kgf / cm2, P, nozzle Diameter, in mm, Dn, and Diameter of the orifice, in mm, Do, on the responses - cold and hot temperatures of outlet air streams (Tc & Th) in °C. Box Behnken design is used for experimentation. The interrelationship between responses and input variables is modelled using Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) separately. Regression equations have been developed for the responses using RSM. ANN modelling is observed giving better prediction results compared to RSM modelling as is evidence that through R-square and Average Absolute Deviation (AAD) for both the models.