SPRAY DRIED YOGURT WATER SORPTION ISOTHERMS PREDICTION USING ARTIFICIAL NEURAL NETWORK

摘要

A predictive model of water sorption isotherms using artificial neural network isrnproposed for spray dried yogurt powders added with hydrocolloids. The model takesrninto account the moisture content, the temperature and the type of gum. Three differentrngums (xanthan 0.1% d.b; k-carrageenan 0.15% d.b; and gellan 0.1% d.b.) were added atrnyogurt samples to improve flow properties. Yogurt was spray dried at the followingrnconditions: air pressure for atomization (1.5 bar); inlet air temperature (180°C to 200°C)rnand outlet air temperature (70°C and 80°C). The static method of desiccator (sulfuricrnacid solutions) was used to determine experimental sorption isotherms at tworntemperatures (30°C and 35°C). The best fit with training data set was found using sixrnneurons, which leave to predict the sorption isotherms with accuracy at least as good asrnthe experimental error over the whole experimental range. On the validation data set,rnsimulations and test experimental sorption isotherms were in good agreement. Thernmodel was compared with two others different model founds in the literature. Therndeveloped model can be used for the prediction of isotherms of spray dried yogurtrnpowder.