Prediction of Heat Transfer Coefficient of Pool Boiling Using Back propagation Neural Network

Abstract

Artificial neural network (ANN), in comparison with empirical correlations, hasrecently received more attention. The present paper includes predictive modeling ofheat transfer coefficient for binary mixtures in pool boiling for hydrocarboncompounds, using Back propagation techniques through Multilayer Perceptron, oneof the types of the artificial neural networks. To train and learn the system, predictiveneural network was found, which is capable of understanding and predicting the presetoutput which is heat transfer coefficient. The principle operation of such neuralnetworks is based on the experimental data collected from some researchers [1-4]. Anew ANN model is proposed using five inputs (mole fraction, temperature difference,heat flux, density and viscosity) to predict the heat transfer coefficient. The predictionusing ANN shows 0.0026 AARE (Absolute Average Relative Error) with most widelyknown correlations namely those of Calus, Fujita and Thome which have given 0.086,0.066 and 0.038 respectively.