AI-Driven Micro Defect Detection for Aerospace-Grade Metal Surface Quality Control in Smart Manufacturing

In the context of Industry 4.0 and Smart Industry, automated and precise quality control systems are essential for maintaining high standards in manufacturing processes. This paper introduces an advanced approach for detecting micro defects on metal surfaces, leveraging the YOLO11 object detection algorithm to meet the stringent demands of modern industrial environments. Micro defects, such as cracks and fractures on metal surfaces, can compromise structural integrity and lead to costly failures, making their detection crucial. In this study, we categorize micro defects into four distinct types, linear, curved, point, and branched, and employ a synthetic defect generation process to create an extensice set of defect variations, subsequently augmented to enhance training diversity. The YOLO11 algorithm is trained and tested across this extensive dataset, achieving high precision and recall rates for each defect category. Detailed precision-recall analysis indicates the model's strengths and areas for improvement, particularly with smaller, more challenging defect types. The results show YOLO11's capability to accurately detect and localize defects, identifying isolated defects, recurring process-related defects, and potential error-prone zones on the metal surfaces. These capabilities align with Industry 4.0's goals of leveraging AI-driven automation to enhance quality control processes, enabling predictive maintenance and operational efficiency.