The dual-channel convolutional neural network for rotating machinery fault diagnosis based on HHT and TMSST

Abstract Fault diagnosis of rotating machinery is crucial for ensuring the reliability of industrial equipment, especially when dealing with nonlinear, non-stationary signals and limited sample data. This paper proposes a novel fault diagnosis method that combines the Hilbert-Huang Transform (HHT) and Time-frequency Manifold Singular Spectrum Transformation (TMSST) with a dual-channel Convolutional Neural Network (CNN). The method utilizes HHT and TMSST to convert raw signals into time-frequency images, which are then processed by the CNN to extract key features and classify fault types. This approach effectively addresses small-sample scenarios, enhancing the accuracy and robustness of rotor fault diagnosis. Additionally, the method incorporates standard deviation analysis to ensure reliable results over multiple runs, demonstrating stability under various conditions.