Reinforcement learning based ball mill discharge size distribution control

The particle size distribution of ball mill grinding products is an important production index of the grinding operation, and due to its multivariate strong coupling and nonlinearity in the operation process, it leads to difficulties in achieving optimal control of the discharge particle size distribution. To this end, a reinforcement learning-based optimal control method is proposed for the control of ball mill discharge size distribution. The proposed method implements the strategy evaluation and strategy iteration using different neural networks and establishes a model-free controller. Finally, the method is compared with the model predictive control method through simulation experiments, and the results show that the proposed method outperforms the predictive control method in terms of time consumption and control accuracy.