Resumen

MILLAN TRUJILLO¹, Félix Rafael  y  OSTOJICH CUEVAS², Zoitza. mass transfer prediction by artificial neural networks in osmotically dehydrated fruits . INCI [online]. 2006, vol.31, n.3, pp.206-210. ISSN 0378-1844.

Summary The purpose of this work was to predict the macroscopic behavior of the two main mass transfer phenomena in three fruits (melon, papaya and apple) osmotically dehydrated. The effect of five process variables was considered: type of food, osmotic solution concentration, fruit size, temperature and process time. The two dependent output variables were water loss and solid gain of the fruits. An artificial neural network model was developed, consisting of an input layer of five neurons and two hidden information processing layers of five neurons each, using sigmoid transfer functions for communication, and two output neurons in order to represent the dependent variables of the model. The neural architecture, trained by the Levenberg-Marquardt algorithm, predicted more than 90% of the data variability for two transfer phenomenon studied, becoming an alternative model to parametric equations developed up to now

Palabras clave : Deshidratación Osmótica; Frutas; Redes Neuronales.