Fault diagnosis of control moment gyroscope based on a new CNN scheme using attention-enhanced convolutional block

Control moment gyroscope (CMG) is a typical attitude control system component for satellites and mobile robots, and the online fault diagnosis of CMG is crucial because it determines the stability and accuracy of the attitude control system. This paper develops a data-driven CMG fault diagnosis scheme based on a new CNN method. In this design, seven types of fault signals are converted into spectrum datasets through short-time Fourier transformation (STFT), and a new CNN network scheme called AECB-CNN is proposed based on attention-enhanced convolutional blocks (AECB). AECB-CNN can achieve high training accuracy for the CMG fault diagnosis datasets under different sliding window parameters. Finally, simulation results indicate that the proposed fault diagnosis method can achieve an accuracy of nearly 95% in 1.28 s and 100% in 2.56 s, respectively.