Heat exchanger modelling using artificial neural networks

This thesis includes estimations of Nusselt number and friction factor using artificial neural networks (ANN) for plate heat exchanger. The feedforward backpropagation algorithm, the most common learning method for ANNs, is used to train and test the network. Data which will be used to train and test the neural network configuration are obtained from three separate studies, one of which is an experimental study performed in the scope of this study. ANN method is first applied to the data taken from the literature and then examined for both the laminar and turbulent flow. ANN predictions are compared with the correlations obtained from the same data. The results are given in detail. Secondly, ANNs are used for the data obtained from experiments which were previously performed at TOBB ETU Heat Exchanger Laboratory. Correlations obtained for three different plates after experimental studies are compared with ANN results and the results are presented in detail. Finally, experimental studies are performed for new plates at TOBB ETU Heat Exchanger Laboratory. After correlations and uncertainty calculations for limited experimental data are obtained, predictions of ANN are compared with the results of correlations developed specifically for the experimental data. After three different studies, it is shown that ANN provides better estimations compared to correlations for property prediction in plate heat exchangers.