A fuzzy system based on multi-objective hierarchical genetic method is proposed to measure the overflow particle size. The fuzzy system are divided into four layers: the input layer, the membership layer, the rule base layer and the system layer. In order to achieve the purpose of co-evolution for each layer, a coding strategy for each layer is designed here. The mean absolute percentage error(MAPE) and root mean square error(RMSE) are considered as the optimization target to calculate the fitness value of each individual. A L-M Bayesian regularization algorithm is used for training the fuzzy system to avoid the ill-conditioned solution. The experimental results using a series of practical production data coming from a grinding plant show the effectiveness of the proposed method.