Using unknown input observers to detect and isolate sensor faults in a turbofan engine

An observer based fault detection and isolation method is presented. The method involves combining the unknown input observer theory with the Beard fault detection filter theory to generate a residual signal that has both disturbance decoupling and unidirectional properties. This robust, in the sense of disturbances, residual was used to detect and isolate sensor faults in a turbofan engine model. Finally, the method was successfully extended to detect and isolate sensor faults over a wide range of power input levels by scheduling the observer based on power code.