Predicting thepH of finished cheese for adaptive control of cheese quality

An adaptive control strategy is proposed to make a better temperature control system during cook stage of an automated cheese production. Adaptive strategies give the alternative temperature set point profiles to the temperature controller, that depend on the pattern of actual acid development status inside a cheese vat and the particular starter culture used for such cheese production. Therefore with an adaptive strategy an automated system has the potential to handle batches of cheese differently.; Artificial Neural Network (ANN) is the “knowledge” part of adaptive control system. It directs the responsive controller to use alternative temperature set point profiles that depend on predicted pH of finished cheese. The results from experiments, comparing with a multiple linear regression model (MLR), showed that ANN can be trained to recognize the pattern of whey pH and temperature change during the cook stage and predict the pH of finished cheese. The successfully trained ANN can give a pH prediction within (0.1) from the observed pH value. This prediction can be made by the first 15 minutes of cook stage.; The MLR model is not able to predict the pH of finished cheese within (0.1) from the observed pH value. However, the model provides the information of variables, which are significantly, explained the predictive pH value. Such variables are the interaction between rate of pH and temperature change during the first 15 minutes of cook stage, the interaction between pH and temperature at 7.5 and 15 minutes cook time and the temperature at 7.5 minutes cook time. The predictive power of ANN is an important element for this control system because the ANN estimation serves as a knowledge component for the system. It allows the controller give to respond by selecting an alternative temperature profile for the rest of the cook stage.; Beside the knowledge from ANN, data acquisition system is a significant piece of the system for both ANN application and adaptive control strategy because it allows on-line data collection during actual cheese production. The system was installed as an additional part of the main cheese vat. This system draws whey solution continuously and measures pH and temperature change during cook stage. Both pH and temperature as a function of time are input variables for ANN to predict the pH of finished cheese.