Abstract: A new fault diagnosis method for stochastic linear control system was presented. When actuator fault occurs, the faults are translated into unknown inputs of the system through mathematic transforms. The actuator fault diagnosis can be realized through using unknown input kalman filter technique. Diagnosis algorithm gives the estimated values of the faults. In order to obtain the accurate values of the faults, the wavelet transform method was adopted. Multi-actuator faults that occur simultaneously or successively can be diagnosed effectively through the organic combination of the two methods. The fault diagnosis system can detect faults in good time and evaluate fault values accurately.