Document Type : Original Article
Authors
1
Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan
2
Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, Iran.
3
Department of Biomedical Engineering, Meybod University, Meybod, Iran.
4
Senior Expert, Technical & Engineering Unit, R&D Department, Sarcheshmeh Copper Complex, Rafsanjan, Iran.
5
Expert, Refinery and Casting Division, Sarcheshmeh Copper Complex, Rafsanjan, Iran.
6
Technical Safety Supervisor of Mines, Health, Safety and Environment (HSE), Sarcheshmeh Copper Complex, Rafsanjan, Iran.
10.22091/jemsc.2026.15913.1378
Abstract
Automated inspection of copper cathode blanks is important in electrorefining production lines, where surface defects can interfere with stripping operations, damage reusable stainless-steel blanks, and increase downtime. This study proposes a hardware–software interactive framework for automated defect detection and robotic sorting under real industrial conditions at the Sarcheshmeh Copper Complex, Iran. The system integrates an industrial camera, a Time-of-Flight distance sensor for cathode positioning verification, real-time monitoring software, deep learning-based inspection, and PLC-based reject command generation. YOLOv8 segmentation is used to isolate defect-relevant regions, while classification is performed using a lightweight MobileNetV4 network enhanced with Local–Global Attention. To address class imbalance, an α-balanced focal loss was employed. The dataset consisted of 5,489 real production-line images divided into training, validation, and test sets. Experimental results achieved 98.5% accuracy, 96.9% F1-score, and 99.2% AUC, demonstrating the effectiveness of the proposed framework for deployment-oriented automated quality control and robotic sorting.
Keywords
Main Subjects
)
Send comment about this article