Aiming at the hydraulic servo position system for asymmetric cylinder rolling mill with nonlinearity, parameter uncertainty and unknown load, a compound control algorithm is designed based on the fuzzy adaptive observer and non singular fast terminal sliding mode in this paper. Firstly, the a backstepping controller is designed based on the non singular fast terminal sliding mode and double power reaching law, and second-order sliding mode filters are constructed to estimate the differential signals of the virtual controls, which can avoid the differential explosion during using the backstepping control effectively. Then, fuzzy adaptive observers are used to approximate the uncertain items in the designed controllers, so as to improve the control precision of the system. Theoretical analysis shows that the closed-loop system is globally asymptotically stable. Finally, simulation is carried out by using the actual data of the hydraulic servo position system of a 650mm reversing cold rolling mill, and the results show that the proposed control method can enhance the asymptotic tracking performance and robust stability of the system in the whole global process through comparing with the conventional sliding mode control method.