인공신경망을 이용한 고장 심각도에 따른 무인기용 유성감속기의 고장진단 연구

Abstract

Purpose: Gear teeth in the rotary geared actuator of the unmanned aerial vehicle, experience crack propagation because of its harsh operating conditions. To prevent the failure of catastrophic events, this study proposes a diagnostic approach for various gear crack levels based on the built-in and add-on sensor signal.<BR>Methods: A downsized planetary gearbox test rig was prepared, in which the motor position, current, and vibration signals were acquired for the normal and 4 different crack-induced states. Signals were filtered around the region of resonance frequencies by spectral kurtosis and the features for the health state were extracted. Then, feature selection was conducted based on the correlation with fault levels. Finally, the Artificial Neural Network (ANN) model was constructed to identify different fault sizes of the cracks, and K-fold validation was adopted to optimize the parameters of the ANN model.<BR>Results: Among the various signals, the vibration from the add-on sensor and a position from the built-in sensor exhibited high performance compared to the current signal. The features after band-pass filtering yielded a high correlation with fault severity.<BR>Conclusion: The proposed method successfully diagnosed different fault severities of gear cracks in the planetary gearbox by using both the built-in and add-on signals.