Predicting moisture content of agricultural products using artificial neural networks

Drying of agricultural products is a significant process to store and use them for various purposes. There are few drying methods in agricultural industry, among them fluidized bed drying is widely employed due to its several advantages over the other methods. The prediction of drying characteristics with a small number of experiments is rather efficient since because of the fact that the drying experiments is time consuming and requires tedious work for a single agricultural product. Therefore, several methods such as deterministic, stochastic, artificial intelligence have been developed in order to predict the drying characteristics based on the experimental data obtained from the lab-scale fluidized bed drying system. In this paper, the artificial neural networks (ANN) method was used to predict the drying characteristics of agricultural products such as hazelnut, bean and chickpea. The ANN was trained using experimental data for three different products through the back propagation algorithm containing double input and single output parameters. The results showed fairly good agreement between predicted results by using ANN and the measured data taken under the same modeling conditions. The mean relative error (MRE) and mean absolute error (MAE) obtained when unknown data were applied to the networks was 3.92 and 0.033, respectively, which is very satisfactory. © 2009 Elsevier Ltd. All rights reserved