Modeling the Cooking Process During the Extraction of Oil from Corn Germ Seeds Using Artificial Neural Networks

In this investigation, to design the process of oil extraction from corn germ on an industrial scale, three cooking temperatures (80, 85 and 90 °C) and three different moisture contents of the output seeds from the cooking pot (3, 3.5 and 4%) were considered and the quantity of insoluble fine partial and oil acidity, oil, protein and moisture contents of the obtained meals were studied as responses. To predict the changes' trend the artificial neural network in MATLAB R2013a software was used. By studying the various networks of back propagation feed forward network with wide range of various topologies, the arrangement of 2-6-5 with a correlation coefficient (R 2 = 0.984) and the mean squared error (MSE=0.003) with using sigmoid hyperbolic of tangent activation function was selected as optimized design. Also Levenberg-Marquardt learning algorithm and learning cycle of 1000 were specified as the best neural model. The results of optimized selected models were evaluated and these models with high correlation coefficients (>0.953) were able to predict the variations process. On the other hand, the results showed that the models obtained in this study had the highest accuracy in predicting the moisture content of the meal. Finally, it was Experimentally determined that for the best properties of corn germ oil and meal, the cooker temperature and moisture contents of the output seeds from the cooking pot should be 80 ° C and 3%, respectively.