Investigation on the use of residual torque as a prognostic indicator for aerospace EMAs

Electromechanical Actuators (EMAs) offer great advantages over their traditional counterparts (namely old Hydromechanical and modern Electrohydraulic Actuators) when used as actuation devices on aircraft. They represent the natural evolution of actuation systems in the more electric and all electric aircraft design philosophies, as using EMAs for both primary and secondary flight controls would eliminate the need for hydraulic and pneumatic power aboard the aircraft, leading to an overall weight reduction and a more convenient way to distribute mechanical power across the aircraft, as distributing electrical power directly to the end users is easier and lighter than distributing pressurized hydraulic fluid. Still, as of today, the use of EMAs is limited to secondary flight control (such as airbrakes, spoilers and high-lift devices) on large aircraft, and they are used as primary flight control actuators only on small UAVs, and, in general, application where the loss of actuation system is neither mission critical nor would lead to loss of life or expensive flying systems. This is partially explained by the fact that EMAs are still a relatively new technology in the aerospace sector: their combined fault modes are yet to be fully understood and they generally lack established prognostic methodologies. Nonetheless, in recent years, many diagnostic and prognostic methods for EMAs have been proposed. The aim of prognostic methods is the estimation of the health status and/or the Remaining Useful Life (RUL) of various components of the EMA so that they can be isolated or replaced accordingly, a cardinal principle of modern Prognostics & Health Management (PHM) philosophies. Many methods proposed to estimate the health status of components rely on the analysis of one or more signals outputted by the system or reconstructed from output variables (as in the case of the back-electromotive force, or BEMF), which are considered prognostic indicators; this approach is often described as hybrid since it leverages both machine learning techniques and knowledge of the physical system. The analysis is thus performed with specifically trained neural networks, which use said prognostic indicators to estimate the health status of one or more components. In this framework, the residual torque, defined as the sum of all the friction and viscous torques in the transmission of the actuator, stands out as a possible candidate to be a valid indicator, as it carries information about the friction coefficients (variation of which are a telling indicator of wearing, possible jamming and other kinds of degradation in the transmission) of the system and can be reconstructed from other data acquired during the functioning of the EMA, such as the electrical current in the motor, the acceleration of the shaft and the hinge moment on the actuator. In this work the viability of the residual torque as a prognostic indicator for EMAs in a neural-network-based methodology is investigated, both in the context of a pre/post-flight routine on ground and of real time use during the flight. The static and dynamic friction coefficients, as well as their ratio and the transmission efficiency under both aiding and opposing loads are considered targets of interest for this application.