ABSTRACT
The present study deals with the prediction of temperature profiles for three liquids namely water, methanol and ethylene glycol at various operating conditions using artificial neural network (ANN) in a natural circulation thermosiphon reboiler. For training of the different networks, the standard feed forward back propagation algorithm was used and several types of structures were tested to obtain the most suitable network for the prediction of temperature profiles. To check the reproducibility of the results, each of the networks studied was trained three times. It was observed that the predicted temperature profiles were very close to the actual experimental data for all three liquids. The predictability of the network is extremely good if the training data are chosen appropriately. If more exhaustive input data are fed: heat flux and submergence then the capability of the network to predict the temperature profile would be better.
In comparison of performance analysis of ANN, the relative error (RE) was studied and maximum error was found to be very low. The training was faster initially then it slowed down asymptotically. The prediction of ANN results was very close to the actual experimental values with a mean absolute relative error less than 2.0 %.
Keywords: Neural networks, Heat flux, Submergence, Temperature profiles, Natural circulation thermosiphon reboiler.
