Open AccessDeep Belief Net-Based Fault Diagnosis of Flight Control System Sensors
Author(s) -
Yuanji Guo,
Wenhui Ning,
Fangyi Wan
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1631/1/012186
Subject(s) - fault detection and isolation , observer (physics) , artificial intelligence , computer science , fault (geology) , control theory (sociology) , deep belief network , identification (biology) , real time computing , artificial neural network , engineering , control engineering , control (management) , physics , botany , quantum mechanics , seismology , actuator , biology , geology
The present study implements a model based on the deep belief net (DBN) into the sensor fault diagnosis of the flight control system. The principle of DBN system identification was adopted to simulate and establish a nonlinear observer of the unmanned aerial vehicle (UAV) for the online estimation of sensors’ output, which identified the fault type by analyzing the residuals of estimated data and the actual output. After the fault detection is completed, the measured value of the faulty sensor is isolated and replaced by the observer-generated value, in order to ensure the UAV normal flight. The proposed DBN model uses the data of normal flight control system sensors as training samples for offline training. A flight control digital simulation system was established to select the optimal DBN model via comparative test runs. Eventually, the common faults of sensors were analyzed and diagnosed online. The results obtained strongly indicate that the proposed method ensures a rapid and accurate diagnostics and isolation of faults, as well as provides a proper signal reconstruction.