Abstract: The diagnosis of helicopter rotor imbalance faults, including mass and aerodynamic imbalance has generally depended on the measurements of the flap and lead /lag track of the tip of the rotor blades and the fuselage vibration at 1/rev frequency. It is of practical significance to simplify the balancing procedures. Based on the authors' previous work, a novel method is proposed for the diagnosis of multiple imbalance faults on the helicopter rotor by using only information extracted from rotor induced fuselage vibrations without any rotor blades track measurements and associated optical devices. It is proved that there exists an injection of rotor imbalance multi-faults space into the rotor induced fuselage vibration space. Using three BP neural networks identifies the injection, one BP neural network identifies the types of the faults, and another two detect the extents of the faults. Those networks are trained and tested by measured fuselage vibration signals in the frequency domain. The location of the unbalanced blade is identified by the initial phase of the fundamental harmonic component of the measured data. Tests are conducted to verify the above approach. The results of the tests show the feasibility and validity of the method.

Key words: helicopter, rotor, fault diagnosis, neural network, frequency domain analysis