Lean Manufacturing Meets AI: Assessing CNN Approaches for Waste Reduction Through Steel Surface Defect Detection

To meet global steel demand in construction, automotive, defense, and machinery, high-quality production is essential. Manual inspection methods are inefficient and subjective, prompting a shift to automated defect detection using Convolutional Neural Networks (CNNs). This study evaluates CNN models (CancerNet, ExceptionNet, InceptionNet, and a custom CNN) on the NEU-CLS-64 dataset, emphasizing metrics like accuracy, precision, sensitivity, and F1 score. MobileNet excels in performance, albeit at a higher cost, while ZFNet lags. Traditional manual defect detection methods are time-consuming and prone to inaccuracies, highlighting the efficiency of CNNs aligned with Lean manufacturing principles. The paper explores various CNN-based image classification algorithms, emphasizing their role in enhancing defect detection in steel production, aligning with Industry 4.0 principles. Decision-makers gain crucial insights for selecting effective CNN models in steel surface detection processes.