A NEURAL NETWORK APPROACH FOR THERMAL PROCESSING APPLICATIONS

The use of a neural network approach in thermal processing applications is presented. A four layer neural network with 3 inputs and 3 outputs was trained using a back‐propagation algorithm. A finite difference computer program was used to predict nodal temperature responses of conduction heating model foods under thermal processing conditions. Equivalent lethality processes were obtained for a range of input variables (can size, food thermal diffusivity and kinetic parameters of quality factors) for sterilization temperatures between 110 and 134C (at 2C intervals). the computed optimum conditions and their associated quality changes were used as input variables for training and evaluation of the neural network. the trained network was found to predict optimal sterilization temperatures with an accuracy of ± 0.5C and other responses with less than 5% associated errors.