Classification of Degree of Degradation around Scribe for Coil-Coated Metallic Samples Using Convolutional Neural Models

Abstract

Coil coating, a technique for applying organic coatings to rolled metal strip substrates, plays a critical role in achieving consistent, high-quality surface finishes. However, these protective coatings are vulnerable to mechanical damage, which can lead to irreversible alterations when exposed to environmental elements. Traditionally, assessing degradation resistance in coil-coated materials involves manual determination of degraded areas. In this study, a classification-based approach to automate this assessment is proposed. Additionally, the selected classification models are compared with semantic segmentation, highlighting their performance and computational efficiency. The results demonstrate that both approaches (classification and semantic segmentation) can assess degradation, with semantic segmentation providing highly accurate results and classification models offering efficient practical deployment alternatives.