Development and experimental validation of prognostic algorithms for electromechanical actuators

In this thesis, the development and construction process of a test bench capable of simulating the operation of an electromechanical actuation system for a secondary flight control surface is analysed and a High-Fidelity numerical model (HF) developed in Simulink is validated with it. The test bench consists of an actuation module, consisting of a sinusoidal brushless motor, power module, microbox PC and control unit developed by Siemens and a planetary gearbox, designed and manufactured entirely within the Politecnico di Torino, using the Fused Deposition Modelling (FDM) additive manufacturing technique. A transmission made with a roller chain connects this first module to the braking module, thanks to which it is possible to apply a braking torque directly to the motor output shaft, in order to simulate the action of external forces. After some initial system tests, it was necessary to modify parameters and elements of the Simulink HF model to obtain the same trends and to ensure better response, a genetic algorithm was used to optimally set seven parameters. After this phase it was possible to validate the Simulink HF model by analysing the behaviour of the test bench with time and frequency domain tests and then repeating these tests on the Simulink HF model. The time domain tests involved position and speed commands. In the case of position control, a step command was applied and the test was performed under two different conditions: with active braking load and with zero braking load. In the case of speed control, a step command was applied and the test was carried out with active braking load and with zero braking load. During the frequency tests, the system was provided with an open loop, sinusoidal position command, evaluating a frequency range extending from 0.1 Hz to 100 Hz for three different amplitudes. Tests in the time and frequency domain were conducted on the test bench and on the Simulink HF model and the latter was able to replicate the signals developed by the former. With this validation of the HF model it is possible to implement prognostic algorithms, evaluating the responses produced by the HF model and by a Low-Fidelity (LF) model, under nominal conditions or in the occurrence of faults (eccentricity and partial phase short circuit), in order to detect discrepancies and identify faults.