Using neural networks to predict thermal conductivity of food as a function of moisture content, temperature and apparent porosity

An artificial neural network (ANN) model is presented for the prediction of thermal conductivity of food as a function of moisture content, temperature and apparent porosity. The food products considered in the present study were apple, pear, cornstarch, raisin, potato, ovalbumin, sucrose, starch, carrot and rice. Thermal conductivity data of food products (0.012-2.350 W/m K) were obtained from the literature for a wide range of moisture content (0.04-0.98 on wet basis, fraction), temperature (-42-130°C) and apparent porosity (0.0-0.70). Several configurations were evaluated while developing the optimal ANN model. The optimal model ANN model consisted two hidden layers with four neurons in each hidden layer. This model was able to predict thermal conductivity with a mean relative error of 12.6%, a mean absolute error of 0.081 W/m K. The model can be incorporated in heat transfer calculations during food processing where moisture, temperature and apparent porosity dependent thermal conductivity values are required. Rahman's model (data considered only above 0 °C) and a simple multiple regression model (all data points) predicted thermal conductivity with mean relative errors of 24.3 and 81.6%, respectively.