Aiming at the servo mechanism fault of launch vehicle, a FDD (fault detection and diagnosis）based on extended multiple model adaptive estimation was proposed. Attitude dynamic model of launch vehicle considering servo mechanism fault was established; the fault angle was used as the state variable to obtain the augmented state space model; the nonlinear estimation of state vectors and fault parameters was carried out by using the extended Kalman filter, and based on the sensor measurement data, the occurrence probability of faults was calculated by the hypothesis testing algorithm; the fault detection and diagnosis procedure based on the extended multiple model adaptive estimation was presented. Simulation results shows that, not only the health monitoring of the servo mechanism can be carried out by the algorithm without fault, but also under single servo mechanism fault, the core servo mechanism whose fault appeared can be timely and exactly detected through the algorithm, and the angle of nozzle under servo mechanism fault can be estimated accurately.