Artificial neural network modeling of heat transfer in a staggered cross-flow tube type heat exchanger

Abstract

This paper presents the application of Artificial Neural Network (ANN) in modeling the heat transfer coefficient of a staggered multi-row, multi-column, cross-flow, tube-type heat exchanger. Heat transfer data were obtained experimentally for air flowing over a bank of copper tubes arranged in staggered configuration with 5 rows and 4 columns at different air flow rates with throttle valve openings at 10 - 100%. The Reynolds number and the row number were used as input parameters, while the Nusselt number was used as output parameter in training and testing of the multi-layered, feed-forward, back-propagation neural networks. The network used in this study was designed using the MATLAB® Neural Network Toolbox. The results show that the accuracy between the neural networks predictions and experimental values was achieved with Mean Absolute Relative Error (MRE) less than 1 and 4% for the training and testing data sets respectively, suggesting the reliability of the networks as a modeling tool for engineers in preliminary design of heat exchangers.