The highly-efficient dry separation technique using a gas-solid fluidized bed is very beneficial for increasing coal grade and optimizing the utilization of coal resources.The size distribution of the solid medium(e.g.,magnetite powder)used in this technique is one of key factors that influences fluidization and separation performance.It is,therefore,urgent to prepare medium in a way that operates at low cost and high efficiency.Grinding experiments were performed using a planetary ball mill equipped with a frequency convener.The effect of fed mass,rotation frequency of the mill,grinding time and the ball-size ratio on grinding performance was investigated.The grinding parameters were optimized by numerical calculations using Artificial Neural Network(ANN)in Matlab.A regression equation for predicting the yield of the desired product(i.e.,0.3～0.15 mm magnetite powder)is proposed.The maximum yield of 0.3～0.15 mm panicles was 47.24%.This lays a foundation for the industrial-scale production of the solid medium required for separation with a magnetite-powder fluidized bed.