Mathematical and neural network modelling of the adsorption characteristics of dried African catfish (clarias gariepenus)

Abstract

Moisture adsorption isotherms of dried fillets of African catfish (Clarias gariepenus) were determined at 23, 30 and 45°C in the range 0.11 to 0.93 of water activity, using the standard gravimetric static method. The data obtained from the experiments were fitted to four popular mathematical models [the Brunauer, Emmett and Teller (BET), Oswin, Henderson and the modified Guggenheim-Anderson-De Boer (GAB)] and to a set of artificial neural network (ANN) models. The modified GAB outperformed the other mathematical models [with coefficient of determination R2 = 0.96, mean relative error (MRE) = 6.64%] but the ANNs were more accurate (with the best giving R2 = 0.988 and MRE = 3.56%) in predicting the moisture adsorption isotherms of the dried catfish. The knowledge of the adsorption isotherms would be very useful in the design of packages, prediction of storage stability and for calculating moisture changes that may occur during storage.