Neural network modeling of heat transfer to liquid particle mixtures in cans subjected to end-over-end processing

Artificial neural network (ANN) modeling was used for the overall heat transfer coefficient (U) and the fluid to particle heat transfer coefficient (h(fp)) associated with liquid particle mixtures, in cans subjected to end-over-end rotation. Both U and h(fp) are needed for modeling the time-temperature profiles of liquid and particles. Experimental data obtained for U and h(fp) under various test conditions were used for both training and evaluation. Multi-layer neural networks with seven input and two output neurons (for a single particle in a can), and six input and two outputs neuron (for multiple particles in a can) were trained. The optimal network was obtained by the varying number of hidden layers, number of neurons in each hidden layer and learning runs, using a back-propagation algorithm. Heat transfer coefficients were also predicted using dimensionless correlations developed earlier from the same data-set. Prediction errors associated with ANN were less than 3 and 5%, respectively, for U and h(fp), which were about 50% better than those associated with dimensionless number models, indicating that the predictive performance of the ANN was far superior than that of dimensionless correlations. The ANN models were also more versatile than the dimensionless number models for predicting U and h(fp). (C) 1997 Canadian Institute of Food Science and Technology. Published by Elsevier Science Ltd.