Abstract: A method of engine sensor fault diagnosis based on geometric pattern recognition technology is proposed to distinguish the sensor slow drifting fault from model mismatch because of install tolerance and performance degeneration. The sensor output is fed to adaptive the performance model, and then a set of component performance modification factors are produced which are regarded as fault patterns. Different faults or performance degeneration produce different patterns which can be distinguished using geometric pattern recognition. Simulation on a turbofan engine shows that the method can diagnose the sensor slow drifting fault, and can monitor the engine health condition.

Key words: aerospace propulsion system, aeroengine, geometric pattern recognition, fault diagnosis, sensor, adaptive model