This paper investigates the implementation of a model-based predictive control (MPC) strategy to improve the performance of a semi-autogenous grinding (SAG) mill in a uranium mineral processing plant. The SAG mill, crucial in crushing and grinding uranium ore to the desired size, is currently managed using conventional proportional-integral-derivative (PID) controllers. However, to enhance production efficiency and control over the SAG mill's variables, this paper suggests the adoption of MPC. The proposed MPC controller is developed using a neural network (NN) model of the SAG mill, created in MATLAB with data collected over 21 days. The effectiveness of the MPC controller is assessed by contrasting its response with that of the real-time operator control. This comparison utilizes tools like MATLAB and the RSlinx remote server for accessing OPC real-time data. Findings reveal that the MPC controller exhibits a quicker reaction to alterations in the SAG mill's process outputs and proficiently regulates crucial outputs such as Mill mass, ensuring that the manipulated variables stay within their designated limits. Unlike operator control, which is slower and adjusts one variable at a time, the MPC approach can maximize the mill's throughput rate without impacting the ore feed rate. This demonstrates the MPC controller's superior ability to optimize SAG mill operations efficiently.