ABSTRACT
This work concentrates on moisture sorption isotherm modeling by the use of GAB equation and artificial neural network (ANN) in legumes. The equilibrium moisture content for adsorption was determined by static gravimetric technique. The experiments for legumes were carried out in saturated salt solutions at 30, 40, and 50 °C until the equilibrium moisture content (EMC) was obtained. The moisture adsorption data obtained were fitted in GAB equations and ANN by using MATLAB. For the ANN modeling seven input neurons corresponding to the seven input variables viz., a w, temperature, ash content, dietary fiber, crude protein, fat content, and carbohydrates were considered whereas the output neuron represents the EMC. The highest R-value and the least mean square error (MSE) value correspond to the hidden neurons of five (for chickpeas) and nine (black-eyed peas, mung beans, and Bengal grams) with the neural network of 7-5-1 and 7-9-1, respectively. At given moisture content, water activity value is estimated by the fitted GAB model in the moisture range from 5 to 20%, at all the temperatures. The generalized isotherm model involving isokinetic temperature and two other parameters, namely, K 1 and K 2 were developed. The overall values were K 1 = 2346237.907 and K 2 = 0.8325. These two parameters were used for the hydration kinetics study of the same product. For the kinetics of hydration of Bengal gram, black-eyed peas, chickpeas, and mung beans at different temperatures, studies were carried out with measurement of moisture content at known intervals considering the principle of one-dimensional mass transfer to the spherical body. For expressing the dependence of the diffusivity value on temperature and moisture, the model involved an expression containing the parameters K 1 and K 2 and heat of sorption. Simulations were carried out to estimate the value of the pre-exponent of diffusivity expression (D o).
