Research on the Method of Industrial Equipment Fault Detection and Identification Based on Improved YOLOv8

Abstract To solve the problems of slow processing speed and misdetection and omission in the presence of cluttered backgrounds in fault detection algorithms, in this paper, combined with the improved YOLO v8 algorithm and the SENet channel attention module, a fault identification and detection method for industrial equipment is proposed. Firstly this method optimizes the network structure of the YOLOv8 algorithm by using the channel attention module SeNet, and effectively capturing the features of key targets and reducing the number of convolution kernels and output feature channels. Secondly, this paper uses data augmentation techniques to enhance the training set and improve the robustness of the YOLOv8 model for small target detection. Finally, the experimental results are analyzed, and the improved YOLOv8n algorithm achieves 96.84% in detection accuracy, 98.73% in recall, and 97.81% in F1-Score, which is excellent in industrial equipment fault detection, and verifies that the YOLOv8n algorithm embedded with the SeNet channel attention mechanism has higher accuracy and stability. Compared with other algorithms, the YOLOv8n algorithm has a greater improvement, and when compared with other industrial detection models, the feasibility of deep learning in equipment fault identification and detection has been verified in three aspects of identification of equipment switch state, identification of abnormal equipment indicator lights, and identification of abnormal equipment display data, demonstrating strong competitiveness.