Modeling Drying Properties of Pistachio Nuts, Squash and Cantaloupe Seeds under Fixed and Fluidized Bed Using Data-Driven Models and Artificial Neural Networks

Abstract

This paper presents the application of feed forward and cascade forward neural networks to model the non-linear behavior of pistachio nut, squash and cantaloupe seeds during drying process. The performance of the feed forward and cascade forward ANNs was compared with those of nonlinear and linear regression models using statistical indices, namely mean square error (MSEMSE), mean absolute error (MAEMAE), standard deviation of mean absolute error (SDMAE) and the correlation coefficient (R2). The best neural network feed forward back-propagation topology for the prediction of effective moisture diffusivity and energy consumption were 3-3-4-2 with the training algorithm of Levenberg-Marquardt (LM). This structure is capable to predict effective moisture diffusivity and specific energy consumption with R2= 0.9677 and 0.9716, respectively and mean-square error (MSEMSE) of 0.00014. Also the highest R2 values to predict the drying rate and moisture ratio were 0.9872 and 0.9944 respectively.