Online framework of prognostic and health management for cmg under multiphysics

Control moment gyroscope (CMG) is the key actuator of satellite and it is critical for monitoring health condition and maintenance. However, there are three challenges in health management. Firstly, the environment where a CMG works is complex and is hard to be decomposed into the receiver and the affected, which brings difficulty to estimate the performance of CMG. Secondly, because of design requirement and installation cost, the vibration sensor, which is regarded as a key health indication other industrial background, is unavailable. Finally, missing data is a big data with unconnected base stations for transforming collected sensor data, decreasing the accuracy of online CMG health condition estimation and rest useful life (RUL) prediction. To face these challenges, this paper proposed a system framework of prognostic and health management considering environment. The contributions in this paper include: 1) introducing a relationship between temperature, current and speed of high-speed bearing to depict the degeneration process. 2) considering the environmental effect on system and designing a duel-flow deep learning model to detect abnormal time serial data. 3) proposing a workflow before data training for estimate missing data. 4) Describing software architecture for spacecraft diagnostics and health management. Finally, experiments and an application example will demonstrate the efficiency of proposed method and framework.