R. Amiri Chayjan; J. Amiri Parian
Volume 36, Issue 3 , December 2013, , Pages 13-24
Abstract
In this research, empirical models and Artificial Neural Networks (ANNs) were used for estimation of equilibrium moisture content of seed and kernel of sunflower. Four empirical models ...
Read More
In this research, empirical models and Artificial Neural Networks (ANNs) were used for estimation of equilibrium moisture content of seed and kernel of sunflower. Four empirical models of modified Henderson, Chung-Pfost, Halsey and GAB were used. Two types of back propagation networks (feed forward and cascade forward) were tested. In order to train input patterns, levenberg-marquardt was used. Temperature and relative humidity limits for the experiments were between 25-40 °C and 11-96%, respectively. The best results for empirical models were obtained for Halsey and GAB models. The best results for applying neural networks were obtained for feed forward network, topology of 3-5-3-1 and threshold function of TANSIG-TANSIG-LOGSIG. With this optimum network, coefficient of determination and mean relative error were 0.9935 and 0.0501, respectively. These results prove the superiority of neural networks compared with empirical models, because besides producing less errors in prediction of equilibrium moisture content, neural also capable of networks are considering the quality index as an input parameter.
