Abstract: A detection method is presented for sensor failures in FADEC engine.In order to avoid that the output of a fault sensor affects other system states through state feadback,we use a bank of Kalman filters in which different combinations of sensor outputs as inputs of the filters are fed to each filter.According to the characteristics of innovation sequences of each filter,the failure can be detected and isolated.The sample frequencis are different in feedback control path and computation of innovation sequences.When a failure has occured in one sensor but before the failure affects other system states,the failure can be detected and isolated effectively.Therefore,the effects of failure on the outputs of other sensors are avoided.A digital simulation has been made for a rotational speed and temperature control system of a twin spool turbine engine.The simulation results show that this failire detection and isolation method is effective.