On-board electromechanical servomechanisms affected by progressive faults: Proposal of a smart GA model-based prognostic approach

The architecture of the actuation systems implemented in aeronautics in order to actuate the flight controls is changing radically and, in the last years, electromechanical actuators (EMAs) are gradually replacing the older type of actuators based on the hydraulic power. Given that some onboard actuators are safety critical, the practice of monitoring and analyzing the system's response (through electrical acquisition) and, then, to provide an evaluation of the evolution of the fault has gradually become an important task of the system engineering. To this purpose, a new discipline was born, called Prognostics, aiming to predict the moment in which a certain component loses its functionality and is not further able to meet desired performances. Given that many prognostic algorithms rely on model-based approaches (e.g. directly comparing the monitor with the real system or using it to identify the fault parameters by means of optimization processes), the design and development of appropriate monitoring models, able to combine simplicity, reduced computational effort and a suitable level of sensitivity and accuracy, becomes a fundamental step of the prognostic process. To this purpose, authors developed a new EMA Monitor Model able to reproduce the dynamic response of the actual system in terms of position, speed and equivalent current, even with the presence of incipient faults. Starting from this Monitor Model, authors propose a new model-based fault detection and identification (FDI) method, based on Genetic Algorithms optimization approach and parallelized calculations, investigating its ability to timely identify symptoms alerting that a component is degrading. The proposed FDI algorithm has been tested on six different progressive failures (dry friction torques and backlash affecting the mechanical transmission, turn to turn short circuit affecting the coils of the three stator phases and rotor eccentricity), simulating progressive faults and evaluating its accuracy. Results showed an adequate robustness and a suitable ability to early identify EMA malfunctions with low risk of false alarms or missed failures.