Modeling of Soybean Snack Roasting by Infrared Heating Using Artificial Neural Network (ANN)

ntroduction: Soybean is recognized as a good source of essential nutrients including protein, oil and several bioactive compounds and soybean has the potential to be used as snack and roasted nut, but most significant factor responsible for such limitation is probably considered as the characteristic flavor of soybean. Raw soybean has beany, bitter and astringent flavors. Therefore to improve its consumption, the particular flavor of raw soybean must be removed. Roasting might be considered as one of the best methods for this object. In this study, the infrared roaster is designed and soybean has been prepared and roasted according to the experimental condition. In this work, an artificial neural network model was developed for modeling of moisture content of soybean snack during infrared roasting. In order to do this, infrared lamp powers of 250, 350 and 450 W, distance between lamp and sample of 4, 7 and 10 cm and roasting time of 30 min were considered as the inputs and the amount of moisture ratio (MR) was estimated as the output. In addition, three different mathematical models were fitted to the experimental data and compared with the ANN model. Based on these results, artificial neural network model for MR with one hidden layer, Sigmoid function as the transfer function, Levenberg-Marquardt method as the learning rule, 4 hidden neurons, 55% for training subset and 25 and 20 percent for each of validation and test subsets respectively had the best over fitting. The determination coefficient (R2) and root mean square error (RMSE) computed for the ANN model were 0.9992 and 0.01099and for the best mathematical model (Two term model) were 0.9776 and 0.02758, respectively. It was concluded that the artificial neural network model satisfied the work better than the mathematical model concerned with soybean snack roasting.