Modeling of Saffron Drying with Oven by Artifical Neural Network

Saffron is the most valuable spice known in the world. Crocin, picrocrocin and safranal are known as saffron color, taste and aroma indices, respectively. Drying is one of the most important steps affecting the final quality of saffron in terms of crocin, picrocrocin and safranal content. In this study, the efficiency of drying process by oven method based on important variables such as temperature, time and thickness of saffron layer was modeled by artificial neural network method. Modeling results of saffron drying process by oven method showed that if crocin changes under momentum learning rule and tangent transfer function with 8 neurons and 25, 55 and 20% of the data were used for evaluation and test training respectively; the coefficient has the highest correlation coefficient (0.914). Whereas for the changes of picrocrocin, the Levenberg learning law and the tangent transfer function in the number of neurons, 12 designed the best networks with 50, 25 and 25% of data for training, evaluation and testing, respectively (R = 0.986). Safranal changes were also predicted by the Levenberg learning law and Sigmoid transfer function in neuron number 8 with 35, 45 and 20% of the data for training, evaluation and testing with appropriate correlation coefficient of 0.981 and predicted by its network.