Identification of a roller screw for diagnosis of flight control actuator

Abstract

The condition based maintenance is an increasing challenge for flight control systems. In this paper, a methodology for the diagnosis of a roller screw in an electromechanical actuator is proposed. As this component is critical, its diagnosis is essential to use it on aircrafts. The methodology is based on the extraction of features by identifying a model of the actuator. First, a specific waveform, made of increasing steps of speed, is run on the actuator. Then, the measurements are processed to reduce the noise and the bias of the different sensors. In order to accelerate the identification, an equivalent point is calculated for each step of the waveform. Then, the identification is realized and the identified parameters are gathered in a feature vector. Finally, a model including backlash and deformation of the stem is used to validate the approach and to generate a set of data. The aging is simulated by making assumptions on the evolution of parameters. Classification is made by using k-Nearest Neighbors (kNN). Performances of the algorithm on this application are evaluated in terms of precision and robustness.