15C-11 |
Using a neural network for pattern recognition of monterey jack cheese |
C. WAISARAYUTT and J. P. Norback. Dept. of Food Science, Univ. of Wisconsin, Madison, Madison, WI 53706-1565 An important feature of an artificial neural network (ANN) is the capability to identify patterns in complex, nonlinear systems. ANN can “learn” these patterns from the provided data. Acid development during cheese production is very important to the quality of cheese and it has a complex nonlinear relationship to the cheese quality. ANN can be used to model such a relationship. The objectives were to develop the useful method for linking acid development during production to the quality of cheese, then evaluate each training strategies for neural networks modeling this link. Time series data of sampled whey temperature and pH values were collected during the cook process for Monterey jack. The data were manipulated for two neural networks, both were three-layer types but different in input and output neuron categories. Both networks determined the optimal number of hidden neurons through trial and error. The training algorithm was back-propagation. The performance of trained networks illustrated their efficiency and generalizability based on the values of MSE (Mean square error), AME (Average mean error), and R2 (coefficient of determination). The numeric ANN contained 63 input nodes (pH and temperature time series and the cook time length) and 2 output nodes of cheese moisture content and pH value. The optimal trained network has both MSE and AME less than 0.17. The pictorial ANN had 1586 input nodes, obtained from graphic transfer and 4 output nodes of acid, “sweet”, firm and weak cheese respectively. This trained network had a higher efficiency than the first one because it had higher testing R2 (0.92) and lower MSE and AME values (0.12 and 0.09, respectively). The pictorial trained network can be used to identify a finished product quality before the entire process will be done. The predicted information will allow the next study of real-time control process of cheese production.
Session 15C, Dairy Foods
|