Fault Diagnosis of Tobacco Packaging Machine Detection using You Only Look Once Version 8

In recent days, the fault diagnosis of tobacco packaging machines is essential to certify the production efficiency, product quality, and reducing interruption in industrial settings. These machines are complex, involving various organized components functioning under high-speed and high-precision conditions, making fault detection and diagnosis a challenging task. Conventional diagnostic methods depend on human intervention and there are challenges such as choosing the suitable network architecture, handling imbalanced fault data, and assuring real-time implication performance remain substantial. This research paper suggests a novel approach by employing robust techniques for fault diagnosis of tobacco packaging machine to detect the quality of the tobacco packaging by accumulating data from the Carton Packing Tape (CPT) dataset. The collected data is preprocessed using an effective technique namely data augmentation, histogram equalization and the Squeeze and Excitation Residual Network (SE-ResNet) is employed to extract features from the embedded data. The extracted images are further classified using the You Only Look Once version 8 (YOLOv8) model. The proposed approach resulted with the F1-score of 94.77%, Precision of 98.56% and Recall of 94.04% when compared to YOLOv7.