MODELING ISOSTERIC HEAT OF BANANA FOAM MAT USING NEURAL NETWORK APPROACH

Abstract

Information on the adsorption isotherm and the thermodynamic properties can assist in optimizing food processing operations such as drying, packaging and storage in the assessment of the quality of food. In this study, an Artificial Neural Network (ANN) was used for modelling the water activity/Equilibrium Relative Humidity (ERH) of banana foam mat under a range of values of the Experimental Equilibrium Moisture Content (EMC) to calculate the isosteric heat of sorption (q_st) by applying the Clausius-Clapeyron equation. The EMC of three dry banana foam samples at different densities of 0.21, 0.26 and 0.30 g/cm³ was determined by a standard gravimetric method over a temperature range of 35-45°C and a relative humidity range of 32-83%. The modified-GAB model best fitted the EMC data. However, the modified-GAB model was not acceptable for predicting the heat sorption behaviour. A negative value of q_st estimated using the modified-GAB equation was found at a moisture content above 0.24 kg/kg d.b., showing the poor fit of the model. A multilayer feed-forward ANN trained by back-propagation algorithms was developed to correlate the output ERH to three exogenous inputs (foam density, EMC and temperature). The developed ANN models could predict the ERH more accurately than the modified-GAB model. The predictions from the ANN models produced R² values higher than 0.97. No negative q_st values were found using the ANN method."