Prediction of heat transfer coefficient during condensation of R134a in inclined tubes using artificial neural network

An artificial neural network model was developed to predict convective heat transfer coefficient (HTC) during condensation of R134a in an inclined smooth tube for the entire range of inclination angles at different saturation temperatures and regardless of flow pattern. The network was designed and trained using a total of 440 experimental data points collected from the literature. Inclination angle, mass flux, saturation temperature and mean vapor quality were used as input variables of multiple layer perceptron (MLP) neural network, while the corresponding HTC was selected as its output variable. By trial-and-error method, MLP network with 18 neurons in the hidden layer was achieved as optimal structure of the ANN which made it possible to predict the HTC with a high accuracy. Mean absolute percent error (NAPE) of 1.48% and correlation coefficient (R) of 0.997 for training data and MAPE of 1.94% and R value of 0.995 for testing data were obtained. Also, 95% and 99% all data were within +/- 5% and +/- 7% error band, respectively. MAPE of 1.61% and R value of 0.9963 were calculated for all data. These results confirm the high ability of the ANNs for predicting the HTC values for the entire range of inclination angles and independent of the flow pattern. (C) 2016 Elsevier Ltd. All rights reserved.