INTELLIGENT AUTOMATED SYSTEM FOR MONITORING DEFECTS IN CONVEYOR BELTS BASED ON NEURAL NETWORK ALGORITHMS OF COMPUTER VISION
DOI:
https://doi.org/10.35546/kntu2078-4481.2026.1.20Keywords:
computer vision, belt conveyor, belt defects, neural network algorithms, ResNet-18, Fast R-CNN, Mask R-CNN, automated diagnostics, segmentation, industrial safetyAbstract
The article addresses a relevant scientific and applied problem of ensuring the reliability of belt conveyors through the implementation of an intelligent automated system for monitoring conveyor belt defects at industrial enterprises. Belt conveyors are critical components of technological processes in the mining, metallurgical, cement, food, and many other industries, and the functioning of entire production lines depends on their continuous operation. At the same time, traditional belt diagnostic methods – primarily visual inspections performed by personnel – have significant limitations associated with the human factor, the need to stop equipment, and the inability to provide continuous monitoring. The aim of this study is to develop and experimentally substantiate the effectiveness of a computer vision system for automated belt defect detection under real production conditions. A comprehensive diagnostic system architecture is proposed, consisting of three levels: a general belt condition classification module (based on ResNet-18), a typical defect detection module (Fast R-CNN), and a damage segmentation module for precise contour extraction and defect area estimation (Mask R-CNN). The models were trained using an experimental image dataset collected from the video stream of an industrial camera. The obtained results demonstrate high fault recognition accuracy and the feasibility of adapting the system to the conditions of Ukrainian enterprises. The classification model shows an error rate below 7 %, while the detector and segmenter provide high-quality defect identification and localization. It is shown that the application of the proposed system enables the transition from reactive maintenance to predictive maintenance, minimizing failures and repair costs. Future work considers integrating the system with MES/ERP platforms to establish a full digital loop for equipment condition management in accordance with the Industry 4.0 concept.
References
Application of machine learning algorithms in real-time monitoring of conveyor belt damage / D. Bzinkowski та ін. Applied sciences. 2024. Т. 14, № 22. С. 10464. URL: https://doi.org/10.3390/app142210464
Computational approaches to Explainable Artificial Intelligence: advances in theory, applications and trends / J. M. Górriz та ін. Information fusion. 2023. Т. 100. С. 101945. URL: https://doi.org/10.1016/j.inffus.2023.101945
Deep learning-based damage detection of mining conveyor belt / M. Zhang та ін. Measurement. 2021. Т. 175. С. 109130. URL: https://doi.org/10.1016/j.measurement.2021.109130
Proactive measures to prevent conveyor belt failures: deep learning-based faster foreign object detection / M. Zhang та ін. Engineering failure analysis. 2022. С. 106653. URL: https://doi.org/10.1016/j.engfailanal.2022.106653
AC-SNGAN: Multi-class data augmentation for damage detection of conveyor belt surface using improved ACGAN / G. Wang та ін. Measurement. 2023. С. 113814. URL: https://doi.org/10.1016/j.measurement.2023.113814
Machine vision based damage detection for conveyor belt safety using Fusion knowledge distillation / X. Guo та ін. Alexandria engineering journal. 2023. Т. 71. С. 161–172. URL: https://doi.org/10.1016/j.aej.2023.03.034
Evaluation of the degree of automation and digitalization using a diagnostic and analysis tool for a methodological implementation of industry 4.0 / F. Martell та ін. Computers & industrial engineering. 2023. С. 109097. URL: https://doi.org/10.1016/j.cie.2023.109097
Хрустальова С., Слюсар А. Використання новітніх логістичних технологій для оптимізації діяльності сучасних підприємств. Радіоелектроніка та молодь у XXI столітті. Т. 2: Конференція «Автоматизовані системи та комп’ютеризовані технології радіоелектронного приладобудування». Харків, Україна, 2024. URL: https://doi.org/10.30837/iyf.asctredb.2024.047
Слюсар А. П., Хрустальова С. В. Методи та алгоритми локалізації RFID-міток: сучасні підходи та перспективи. Виробництво & мехатронні системи 2024: матеріали VIII-ої Міжнар. конф., м. Kharkiv, 25–26 жовт. 2024 р. Kharkiv, 2024. С. 87–91.
Application of bayes theory model to determine the optimal variant of technological process for production of MEMS components / I. Nevliudov та ін. 2024 IEEE 19th international conference on the perspective technologies and methods in MEMS design (MEMSTECH), м. Zozuli, Ukraine, 16–19 трав. 2024 р. 2024. С. 24–28. URL: https://doi.org/10.1109/memstech63437.2024.10620011
