A classification method for bearing surface defects based on acoustic emission technology and the YOLO-V11 algorithm

In contemporary petrochemical manufacturing, the identification of defects in a substantial quantity of bearings is frequently a necessity. Conventional methods of detection, namely manual inspection and acoustic emission detection, are often plagued by deficiencies in terms of efficiency and precision. The proposed methodology integrates acoustic emission technology with the open-source deep learning algorithm YOLO-V11, facilitating rapid detection of bearing faults. Initially, five types of bearings of the same model but different varieties are selected and installed on the same shaft segment, which is then connected to an acoustic emission detection device. Acoustic emission signals are obtained for each type of bearing according to the different types of bearing fault. These signals are then visualised in two dimensions to generate vibration images, which serve as input for the model and are used to train the YOLO-V11 model. The experimental findings demonstrate that the prediction accuracy and recall rate for various defects generally exceed 80%, thus substantiating the efficacy of the proposed method for industrial production in the diagnosis and classification of bearing defects.