Moisture sorption isotherms and thermodynamic properties of safflower seed using empirical and neural network models

Abstract

Experimental work was conducted to obtain the adsorption isotherms of safflower seed using a semi static gravimetric method at 25, 40, and 60 °C over a water activity range from 0.11 to 0.94. Fourteen isotherm equations and multilayer artificial neural network approach were employed to analyze the experimental data. In order to evaluate the goodness of fit of each model determination coefficient, Chi Square and root mean square error were used. The Peleg (at 25 and 40 °C) and Osvin (at 60 °C) models were selected to best describe the sorption isotherms of safflower seed. Accordingly, the results showed that the predicted values of ANN model were more accurate than those predicted by nonlinear regression method. The adsorption monolayer moisture content was also determined using Brunauer, Emmett and Teller equation. The monolayer moisture content values were 0.028, 0.024, and 0.022 gH2O/g solid at 25, 40, and 60 °C, and the corresponding constant values of the Brunauer, Emmett and Teller equation were found to be −13.029, −10.873, and −17.255, respectively. The safflower seeds heat of sorption values was also found. The heat of sorption values was large at low moisture content and decreased with an increase in the moisture content, whereas the magnitudes increase with increases in temperature and could be well adjusted by an exponential relationship. The experimental data revealed that enthalpy-entropy compensation theory was satisfactorily applicable to the moisture sorption behavior of the safflower seed.