AUTOMATED METALLIC SURFACE FLAW INSPECTION USING ARTIFICIAL INTELLIGENCE TECHNIQUE

The growing demand for superior metallic components in several industries has emphasized the necessity for effective and dependable inspection techniques. Conventional manual inspection procedures are lengthy, based on personal judgment and susceptible to human mistakes. In this study, we introduce a dynamic multi-layered auto-encoder with a robust deep neural network (DMAE+DNN) system for inspecting flaws in metallic surfaces. We acquired images of the metal surface defects. The neural network design is improved by incorporating a dynamic multi-layered auto-encoder, enabling the system to obtain highly detailed representations of surface data. The results demonstrate the improved performance of the suggested system, showcasing higher levels of recall, precision and F1-score in comparison to conventional defect detection methods. This technological development has the possibility to completely transform quality control procedures by minimizing the need for manual inspections and improving the overall quality of products that heavily rely on metal elements.