Application of Radial Basis Function Network with a Gaussian Function of Artificial Neural Networks in Osmo-dehydration of Plant Materials

摘要

The study presents a critical evaluation of Artificial Neural Networks (ANNs) in food processing by successfully predicting the mass transfer in three plant materials. The used of ANNs in osmo-dehydration was evaluated using two varieties of apple ( Malus domestica Borkh) of Golden Delicious and Cox, banana cultivar Cavendish and potato ( Solanum tuberosum L.) variety Estima . In the ANNs, the radial basis function (RBF) network with a Gaussian function employing the orthogonal least square (OLS) learning method was used. A single hidden layer of few neurones (NHL = 20) resulted in the neural network being limited in its ability to model the process efficiently and the coefficient of determination (R²) was 0.76 for water loss. Increased neurones (NHL = 100) the network was improved significantly (R² = 0.84) for water loss. Subsequent increase of the neurones to 120 (NHL = 120) showed a significant improvement of the network (R² = 0.91) for sucrose gain. The mass transfer in the three plant materials were successfully predicted by the ANN models indicating the ability of ANN to model both linear and non-linear models as an advantage over empirical equations for quality predictions in food processing.