S-ST-1-17 Using Neural Networks for Predicting Nusselt Number for Intube Laminar Flows

P. Kazemi

Neural networks can be used to predict the Nusselt number for intube laminar flow. In addition to describing the architecture of a neural network, this report discusses the application of a neural network to a manufactured data set, open literature data, HTRI data, and CFD simulation results. Our results illustrate the effect of noise on the training process and the analysis of accuracy. We also compare the predictive accuracy of the neural network model to correlations used to predict the Nusselt number and conclude that a neural network can perform as well as correlations given reliable training data.