Neural Network-Based Prediction of Adiabatic Capillary Tube Characteristics With Alternative Refrigerant

Abstract

Capillary tubes are expansion device which are commonly used in small vapour compression refrigeration systems. An artificial neural network (ANN), a feed-forward network with back propagation algorithm was developed using alternative refrigerant. The feed-forward ANN with three-layers is used as a universal approximator for the multiple-input and single-output non-linear function. For the ANN training and testing, experimental data for refrigerant HC290/HC600a/HFC407C mixture from open literature was used. The most appropriate number of neurons in the hidden layer was found to be fifteen with average error between the predicted mass flow rate from ANN and experimental data is less than 1.6%. This paper shows the suitability of ANN for the predictions of mass flow rate of alternative refrigerant flowing through adiabatic capillary tube.