Fault Detection and Identification of spacecraft reaction wheels using Autoregressive Moving Average model and neural networks

Abstract

Spacecraft Attitude Determination and Control System (ADCS) is considered to be one of the most critical subsystem of the low earth orbit satellites due to the pointing accuracy required during its operation. Consequently a fast and reliable Fault Detection and Identification (FDI) technique is obtaining more significant weight meanwhile years of researches. This paper presents a procedure to ameliorate and amend the (FDI) of a spacecraft reaction wheel as a part of the (ADCS) by differentiating the signatures of possible faults which could be occurred inside the reaction wheel such as over voltage, under voltage, current loss, temperature increase, and hybrid faults using Autoregressive Moving Average (ARMA) model for either normal and faulty data based on the behavior of a dynamic mathematical model of 3-axis spacecraft reaction wheel and neural network classifier. The results demonstrate that the fault detection and identification are successfully accomplished.