An application of neural network and genetic algorithm for optimizing food extrusion process parameters

Optimizing process conditions in food extrusion is extremely difficult due to the involvement of multiple process variables and lack of knowledge about their interactions. Previous work provided limited results for simplified scenarios based on unrealistic assumptions. In this study, a more effective approach is proposed. A neural network was developed and trained to map highly nonlinear relationship between input variables and process output. A genetic algorithm based system is then developed to search the best set of parameter values. The neural network was integrated into this system and used as a function estimator to evaluate fitness values of solutions in each generation. The computational experiments have been conducted to validate the developed system and evaluate and compare its performance with traditional regression approach.