Supervisory algorithm based on reaction wheel modelling and spectrum analysis for detection and classification of electromechanical faults

摘要

This study presents the design of a supervisory software algorithm that can detect and classify different types of electromechanical faults and determine the fault source in the reaction wheel. Unlike conventional methods, which are generally based on the modelling of the entire control system, the fault occurrences and all critical points are predicted at the module level. For this purpose, a new model of the wheel is proposed in which all internal interactions are considered. Based on this model, a method of parameter estimation is proposed to classify stator and bearing faults. Furthermore, an innovative approach based on the spectral analysis of the stator current signal is applied to detect and classify faults caused by damage to the permanent magnet or flywheel. These strategies are combined to ensure that all of the electromechanical faults that are likely to occur in the wheel will be detected. Simulation and experimental results obtained from a hardware-in-a-loop test bed and xPC Target toolbox demonstrate the applicability of the proposed algorithms in actual systems.