Optimization of the sheet metal inspection process using artificial intelligence
Erik de Souza Couto, Luis Fernando de Almeida, Wesley Rossi Pimenta, Daniel Costa Trocades, Regis Fonseca de Oliveira, Yuri Lucas de Souza Paropat, Maycon Lacerda de Oliveira, Silvio de Carvalho Sabença
Abstract
This study details the implementation of a computer vision system for automating quality control in the steel industry, with a focus on detecting scratches on steel plates at the CSN cold rolling mill. The objective was to validate the feasibility of an artificial intelligence model as a replacement for manual inspection, which is subjective and prone to error. The methodology involved fine-tuning a model using a custom dataset collected directly from the production process. the system’s viability as a proof of concept, achieving a mAP@50 of 0.574, an F1-Score of 0.584, and an inference time of just 3.2 ms per image, indicating the possibility for real-time application. It is concluded that the solution is technically feasible and efficient, indicating potential contributions to quality control accuracy, reduced operational costs, and the generation of data for continuous process optimization.
Keywords
References
1 Wang J, Ma Y, Zhang L, Gao RX, Wu D. Deep learning for smart manufacturing: methods and applications. Journal of Manufacturing Systems. 2018;48:144-156.
2 Nunes JBL, Cordeiro NH. Aplicação de técnicas de visão computacional para inspeção de defeitos na superfície de chapas laminadas de aço no processo produtivo industrial [monografia]. Votuporanga: Instituto Federal de Educação, Ciência e Tecnologia de São Paulo; 2022.
3 Tudeschini RB, Soares AMS. Automatic inspection system of adhesive on vehicle windshield using computational vision. Journal of the Brazilian Society of Mechanical Sciences and Engineering. 2023;45:124.
4 Boff VC. Sistema de visão computacional para medição sem contato das dimensões de capacitores [monografia]. Caxias do Sul: Universidade de Caxias do Sul; 2020.
5 Tao X, Zhang D, Ma W, Liu X, Xu D. Automatic metallic surface defect detection and recognition with convolutional neural networks. Applied Sciences. 2018;8(9):1575.
6 Sakuma DT. Implementação de um framework de visão computacional para inspeção de qualidade no assoalho de automóveis [dissertação]. São Paulo: Universidade Federal de São Paulo; 2023.
7 Fernandes RR. Inspeção visual em linha de produção utilizando visão computacional [monografia]. São Paulo: Universidade de São Paulo; 2024.
8 Souza KS. Implementação de um sistema de visão com deep learning para otimizar o processo de inspeção de emendas das bobinas na fabricação do cinturão de segurança [dissertação]. Manaus: Instituto de Tecnologia e Educação Galileo da Amazônia; 2024.
9 Pan SJ, Yang Q. A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering. 2010;22(10):1345-1359.
10 Xu L, Xu EL, Li LX. Industry 4.0: state of the art and future trends. International Journal of Production Research. 2018;56(8):2941-2962.
11 Loshchilov I, Hutter F. Decoupled weight decay regularization. arXiv preprint. 2017;1711.05101.
12 Jocher G, Chaurasia A, Qiu J. Ultralytics YOLOv8 [Internet]. 2023 [cited 2025 Oct 9]. Available at: https://github.com/ultralytics/ultralytics
13 Terven J, Córdova-Esparza DM. A comprehensive review of YOLO architectures in computer vision: From YOLOv1 to YOLOv8 and YOLO-NAS. Mach Learn Knowl Extr. 2023;5(4):1680-1716.
14 Lin TY, Maire M, Belongie S, Hays J, Perona P, Ramanan D, et al. Microsoft COCO: Common objects in context. In: Fleet D, Pajdla T, Schiele B, Tuytelaars T, editors. Computer Vision – ECCV 2014. Cham: Springer; 2014. p. 740-755.
15 He Y, Song K, Meng Q, Yan Y. An end-to-end steel surface defect detection approach via fusing multiple hierarchical features. IEEE Transactions on Instrumentation and Measurement. 2020;69(4):1493-1504.
16 Luo Q, Fang X, Liu L, Yang C, Sun Y. Automated visual defect detection for flat steel surface: A survey. IEEE Transactions on Instrumentation and Measurement. 2020;69(3):626-644.
17 Ren S, He K, Girshick R, Sun J. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2017;39(6):1137-1149.
18 Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu CY, et al. SSD: Single shot multibox detector. In: Leibe B, Matas J, Sebe N, Welling M, editors. Computer Vision – ECCV 2016. Cham: Springer; 2016. p. 21-37.
Submitted date:
10/09/2025
Accepted date:
02/18/2026
Facebook
Google+
Twitter
LinkedIn
Mendeley
StumbleUpon
CiteULike
Reddit
Email