88C-12 |
Prediction of thermal conductivity of bakery products using neural networks |
M. MARCOTTE1, S. S. Sablani2, and O. D. Baik1. (1) Food Research & Development Centre, Agriculture & Agri-Food Canada, 3600 Cassavant Blvd. W., Sainte-Hyacinthe, QC J2S 3E8, Canada, (2) Dept. of Bioresource & Agricultural Engineering, Sultan Qaboos Univ., PO Box 34, Al-Khod, Muscat, 123, Oman Several models were proposed to predict the thermal conductivity of foods these but none of them can be used for a wide range. Artificial neural network (ANN) modeling is becoming an interesting method for the prediction of thermal properties. The objective was to develop neural network models for the prediction of thermal conductivity of bakery products as a function of moisture content, temperature and bulk density. Feed forward neural networks were used. The input layer consisted of 3 neurons representing the product moisture content, temperature and bulk density, while the output layer had one neuron for the thermal conductivity. The number of hidden layers varied from one to two and the neurons within, from two to 16. The data set (83 cases) was divided in two. It consisted of 66 cases for training and 17 for validation, chosen randomly. The back-propagation algorithm was used in model training and a hyperbolic-tangent transfer function was used. The random number seed was constant before each training and the learning coefficient ratio was 1.0. The optimal configuration was obtained upon minimizing the difference between predicted and desired values. The performance was calculated using the mean relative error, the mean absolute error, the standard error and the Rē of the linear regression. The optimal ANN model involved a network with six neurons in each of two hidden layers. The optimal ANN model is capable of predicting the thermal conductivity values of various bakery products for a wide range of conditions with mean relative error of about 10%, mean absolute error of less than 0.02 W/m K and standard error of about 0.003 W/m K. The developed model can be used to estimate thermal conductivities of different bakery products for a wide range of conditions.
Session 88C, Food Engineering: Physical and Chemical Properties
|