Modelling moisture diffusivity of pomegranate seed cultivars under fixed, semi fluidized and fluidized bed using mathematical and neural network methods.

Abstract

BACKGROUND: Modelling moisture diffusivity of pomegranate cultivars is considered to be a major aspect of the drying process optimization. Its goal is mainly to apply the optimum drying method and conditions in which the final product meets the required standards. Temperature is the major parameter which affects the moisture diffusivity. This parameter is not equal for different cultivars of pomegranate. So modelling of moisture diffusivity is important in designing, optimizing and adjusting the dryer system. MATERIAL AND METHODS: This research studied thin layer drying of three cultivars of pomegranate seeds (Alak, Siah and Malas) under fixed, semi fluidized and fluidized bed conditions. Drying process of samples was implemented at 50, 60, 70 and 80°C air temperature levels. Second law of Fick in diffusion was utilized to compute the effective moisture diffusivity (D(eff)) of the seeds. Linear and artificial neural networks (ANNs) also were used to model D(eff) of seeds. RESULTS: Maximum and minimum values of the D(eff) were related to Malas and Alak cultivars, respectively. Three linear models were found to fit the experimental data with average R2 = 0.9350, 0.9320 and 0.9400 for Alak, Siah and Malas cultivars, respectively. The best results for neural network were related to feed forward neural network with training algorithm of Levenberg-Marquardt was appertained to the topology of 3-4-3-1 and threshold function of LOGSIG. By the use of this structure, R2 = 0.9972 was determined. CONCLUSION: A direct relationship was found between D(eff) and thickness of fleshy section of the seeds. The Siah cultivar has the highest value of D(eff). This is due to higher volume of fleshy section of the Siah cultivar. Cultivar type and air velocity have the highest and the least effect on D(eff), respectively.