Steel Surface Defect Detection Using Machine Learning Techniques

Data-driven fault diagnosis in steel sheets is an over-due technology in the steel manufacturing industry. Automated identification and categorization of defects will enhance product quality and improve manufacturing efficiency by limiting human error. Existing classifiers utilize supplementary feature extraction procedures to enhance accuracy and are equally effective as the procedure used. The preprocessing parameters are static, unable to adapt to respective defect types, leading to mediocre accuracy. As a result, the current inspection mechanism has scope for improvement. Therefore, this article introduces a deep convolutional neural network for detecting and categorizing damaged steel. The presented network utilizes 18 layers, framed like a VGG-16 architecture. The suggested approach attains an accuracy of 98.89%, which is better than existing methods. The results prove the framework's efficiency, paving the way for real-time examination of surface-level defects.