Application of GA-Optimized ANN on Modeling the Performance of Coiled Adiabatic Capillary Tubes

Abstract

An Artificial Neural Network (ANN) model optimized with Genetic Algorithms (GA) is applied to predict the mass flow rate in coiled adiabatic capillary tubes. Capillary tubes are the key flow control devices in small refrigeration and air conditioning units, which are usually coiled to save space. The flashing flow through coiled capillary tubes is much complex and the physical process is typically non-linear, which needs complicated mathematical model (conservative equations) for precise prediction. A GA-optimized ANN model is thus employed to address this challenging problem, which is valuable for the design of coiled capillary tubes in real applications. The training samples are from the experimental data on a one-pass-through test facility, which provides accurate source datasets. The results show that the predicted mass flow rates with GA-optimized ANN model agree well with the test data with an average error of 2.43%.