An actuator failure diagnosis algorithm was designed based on multi-model parameter estimation to monitor the working state online and present the failure type as well as failure degree in real time.Firstly,the dynamic characteristics of the typical actuator failures were described and the corresponding models were built up for each one.Then some state observers and parameter estimators were designed for each kind of the representative failures to monitor the operating state of the actuators and present the failure information in time.A switching mechanism was used to choose the correct observer for the current failure.The simulation result shows that the failure models we established can describe the dynamic behavior of the fault actuator very well,and the switching scheme can switch to the right observer correctly in time to supply the most accurate failure information.