Circulating fluidized bed (CFB) boiler is a highly efficient and clean combustion technology with a wide range of applications, but the wear problem has always plagued the long-term operation of the CFB boiler. At present, the wear conditions in different areas are mostly master through operational survey experience or by using numerical simulations to obtain velocity and concentration fields，few scholars have studied quantitatively the wear conditions of different areas in the boiler through theoretical methods. The wear is mainly affected by the velocity and concentration of dust airflow. An attempt was made to obtain the velocity and concentration of fly ash particles in the vicinity of the water-cooled wall heating surface under 50 sets of operating conditions using hydrodynamic software simulations in this study, the relative wear prediction model of GA-BP neural network with the structure of 5-13-12 was established by BP neural network and genetic algorithm (GA) for the CFB boiler mixed with petroleum coke in a petrochemical plant. In turn, the effects of five operating parameters, namely the air volume of the air distribution plate, the primary air volume, the secondary air volume, the fuel volume, and the blending ratio, on the wear in different areas of the furnace chamber were analyzed. The results show that the prediction results of the test set are in good agreement with the wear conditions surveyed on-site, which verifies the feasibility of using GA-BP neural network to establish a wear prediction model and can guide the anti-wear operation, under the condition of ensuring the normal operation of the boiler, appropriately reducing a certain amount of fluidized air of the air distribution plate, reducing the amount of primary and secondary air and fuel in the dense phase area, and appropriately increasing the petroleum coke blending ratio, which can reduce the wear of the heated surface in the furnace.