A Random Forest and Model-Based Hybrid Method of Fault Diagnosis for Satellite Attitude Control Systems

Fault diagnosis is the key technology to guarantee the reliability and safety of satellite attitude control systems (ACSs). Although model-based methods have achieved good fault diagnosis performance, various factors (such as closed-loop design, model nonlinearity, and external disturbances) in the satellite ACSs still bring challenges to fault isolation. Meanwhile, data-driven methods can assist model-based methods for fault diagnosis based on residual signals (generated from model-based methods) and input-output data to relieve the difficulty. Based on the above motivations, a novel model and data dual-driven fault diagnosis approach are proposed in this article for satellite ACSs. Firstly, an H_{i}/H_{\infty } optimization-based fault detection filter is considered as a residual generator, which is designed to be robust against disturbance and sensitive to a fault. Then, the occurrence of a fault can be detected based on the residual evaluation. Eventually, a random forest (RF) algorithm is developed to achieve fault isolation with system input-output and residual signals. A simulation experiment, including fault detection and fault isolation of single faults and multiple faults, is conducted to show that the proposed approach is effective and better than the other three methods.