Q-Learning Based Reinforcement Learning Controller for Concentration Control of Food Preparation

Food preservation is the practice of preparing food in such a way that its beneficial characteristics (sensory and nutritional) are maintained for as long as feasible after preparation. Water elimination and reduction of water content are the basis of the food preservation theory. Maintaining concentration at a particular level is crucial in the food manufacturing industry. Common processing equipment in the food industry is the Continuous Stirred Tank Reactor (CS TR). The CS TR's function is to keep the reactor's internal temperature and product concentration constant. This study aims to keep the product concentration constant using two different controllers. The first is a traditional Proportional Integral Derivative (PID) controller, and the second is a Reinforcement Learning (RL) controller based on Q-learning. A mathematical model of the CS TR is constructed, and the controllers are developed. The concentration of the product is set to the specific value and allows the controller to track it. Error measures such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) are used to assess the performance of each controller. The RL controller shows promising results in terms of error metrics, suggesting that it may be a viable option for maintaining concentration in food industries.