Electro-Mechanic actuators are usually driven by Brushless-DC motors. Those ones can be affected by damages caused by fatigue and usage. At incipient stage, failures like partial short-circuit and rotor static eccentricity do not influences overall actuator performance, causing however oscillations in speed and command signals. Focusing on these oscillations, which presents periodic behaviour over the rotor revolution, neural networks are designed and trained to detect and quantify the damage entity. A different approach to this problem is performed, ignoring electrical measures of phase currents and potential drops, which are usually noisy and hardly available as actuator output. Classification results shows good performance in every speed-torque combination, reducing undetected failures and being capable to distinguish between the two kids of failures implemented.
|Titolo:||Neural Network Design for Incipient Failure Detection on Aircraft EM Actuator|
|Data di pubblicazione:||2016|
|Appare nelle tipologie:||1.1 Articolo in rivista|