Mill is acknowledged as main crushing equipment in the ore dressing and smelting process. In addition, the mill load is considered as one of the critical parameters during the operation of crushing equipments. In light of complexity, nonlinearity and uncertainty of the mill load control, an improved neural network control strategy based on chaotic particle swarm optimization was proposed. To realize the quick and accurate control of mill load, the neural network weight was optimized using the chaotic particle swarm, furthermore, the slower BP neural network convergence rate and larger probability to fall into minimum value were improved. Ultimately, the simulation results indicated that the control performance such as stability of mill load control was obviously improved using the proposed strategy.