Accurate fault diagnosis of rotating machinery components is the key to ensuring the safe operation of the mechanical system. Aiming at problems such as inaccurate detection of small target fault features and loss of fault information in the process of advanced feature extraction that exist in conventional machine learning and traditional deep learning methods in the fault diagnosis of rotating machinery, this paper improves the YOLO v8. It proposes a YOLO v8-C-OD fault diagnosis method. Firstly, the original vibration signal is transformed into a time-frequency image with enhanced fault features using continuous wavelet transform (CWT) based on Morlet wavelet as the optimal wavelet basis (WBF) to form an experimental sample. Secondly, the C2F module in YOLO v8 Backbone is improved by introducing Omni-dimensional Dynamic Convolution (ODConv) into C2F to dynamically adjust the weight of each convolution kernel to enhance the fault feature extraction capability. Finally, the Convolutional Block Attention Module (CBAM) is fused into YOLO v8 to suppress unnecessary features and significantly improve the model fault diagnosis rate. Finally, experiments conducted on datasets such as those from Case Western Reserve University (CWRU) and Jiangsu Qianpeng Diagnostic Engineering Co. show that the fault classification diagnosis accuracy of this method reaches 100% and 99.75%, respectively, which outperforms existing state-of-the-art models. This study introduces a novel fault diagnosis method and provides a valuable reference for the fault diagnosis of aircraft and industrial rotating machinery components.
Wang et al. (2025) studied this question.