Skip to Main Content
This paper proposes an online fault detecting and isolating (FDI) scheme of satellite attitude control system (SACS) based on Wavelet and Dynamic Recurrent Neural Network (DRNN) which is capable of processing time-varying signals in real time. First, a novel improved wavelet method is proposed to detect faults; then, a DRNN is designed for fault isolating (FI) and the relevant fault decision module as well. The DRNN is trained by corresponding target FDI result of fault data set sampled from actuator and sensor outputs. Unlike many previous wavelet-based fault detecting methods developed in the literature, our proposed FDI scheme is only based on measurement signals which can avoid modeling, also wavelet method is improved and suitable for online signal processing. Real-time simulation is performed and the results demonstrate the validity and superiority of our method.