Intelligent Fault Diagnosis for Inertial Measurement Unit through Deep Residual Convolutional Neural Network and Short-Time Fourier Transform

An Inertial Measurement Unit (IMU) is a significant component of spacecraft, and its fault diagnosis results directly affect the spacecraft’s stability reliability. In recent years, deep learning-based methods have made great achievements; however, some problems such as how to extract effective features promote training process networks are still be solved. Therefore, in this study, novel intelligent approach combining residual convolutional neural network (CNN) data preprocessing algorithm proposed. Firstly, short-time Fourier transform (STFT) adopted raw time domain into time–frequency images so useful information can extracted. Then, Z-score normalization augmentation strategies both explored exploited facilitate subsequent model. Furthermore, modified CNN-based model, which utilizes Parameter Rectified Linear (PReLU) activation functions blocks, automatically learns classifies types. Finally, experiment’s indicate that proposed method has good features’ extraction ability performs better than other baseline models terms classification accuracy.