Abstract: Modeling the ore grinding and classification process is very important to the optimization of production index. A hybrid intelligent model is thus developed for the simulation of ore grinding and classification process according to its compositive complexity including nonlinearity, multivariable, time-varying parameters and boundary condition fluctuation. This dynamic model is based on the population balance models for ball mill and sump, TSK (Takagi-Sugeno-Kang) model for net mill power draft, empirical cyclone model together with the RBF networks to compensate for the estimation error of overflow density and particle size distribution. With the actual process data from a grinding circuit of a concentration plant, the simulation results of this hybrid intelligent model have the same dynamic characteristics during the variation in fresh slurry feed velocity, density, particle size distribution and cyclone feed manipulating variables.