Abstract: Aiming at the deficiency of bad stability of filter and low precision estimation in the method for sensor fault diagnosis based on strong tracking Kalman filtering，two-filters method is put forward.If the measurement noise variance and system noise variance of a filter are both more than its real noise variance，then its fault estimation precision will be low，but tracking speed will be high；inversely，if only the measurement noise variance of another filter is more than its real noise variance，then its fault estimation precision will be high，and tracking speed will be low.So the first filter can be used to detect the fault timely and the second filter can be used to estimate the fault amplitude accurately.The results of simulation experiments show that this method is effective.