A Three-Stage Data-Driven Approach for Determining Reaction Wheels’ Remaining Useful Life Using Long Short-Term Memory

Reaction wheels are widely used in the attitude control system of small satellites. Unfortunately, reaction failure restricts efficacy a satellite, and it is one many reasons leading to premature abandonment This study observes measurable parameter faulty wheel induced with incipient fault estimate remaining useful life wheels. We achieve this goal three stages, as none observable parameters directly related health wheel. In first stage, we identify necessary predict future these using sensor acquired data long short-term memory recurrent neural network. second index multivariate third based on historical parameter. Normalized root mean squared error evaluate performance various models each stage. Additionally, different timespans (short, moderate, extended scale satellite orbit times) simulated tested for proposed methodology regarding malfunction Furthermore, robustness method missing values, input frequency, noise studied. The results show promising scheme accuracy predicting around 0.01–0.02 normalized error, prediction RUL 1%–2.5%, uncertainty factors, discussed above.