DOI: 10.18503/1995-2732-2025-23-3-122-128
Abstract
With the advent of new tools and technologies in the modern world, specialists in the field of mechanical engineering have more and more technical opportunities to realize their professional tasks. One of these tools is neural networks, and in particular those that can recognize objects in videos and images. The main advantages of using them are a significant increase in labor productivity and a reduction in the number of errors during the work process. In our opinion, one of the best representatives of neural network-based object recognition models is the YOLOv5 model but, unfortunately, there was no information about the possibility of using it to work with drawings, in particular, the detec-tion of details obtained by turning methods. The objectives of the work described below were: firstly, to test the YOLOv5's ability to identify body parts with different ratios of length L to diameter D in graphical images (drawings) and form them into three groups, and secondly, to test the ability to ensure high accuracy and speed of operation when solving the problem of detecting these parts. In the course of the work, two main methods were used: the method of training a model for recognizing objects based on a neural network and the method of working with a set of graphic images (drawings) using the resulting model. As a result, based on the first method, a new model was obtained that can recognize turning parts in graphic images, and further based on the second method, its performance was tested. After the tests, we can definitely say that the resulting model can effectively solve the problem of finding turning parts in drawings, it is also worth noting that the characteristics of the trained model obtained during the work are very high, and based on this, we can say that the model is quite promising for working with drawings. The results obtained can help automate the process of classifying parts by determining their geometric characteristics, as well as open up new perspectives for the use of grouping parts.
Keywords
mechanical processing, turning, parts such as bodies rotation, object detection, neural networks, YOLOv5
For citation
Kuznetsov S.V., Rogovik A.A., Kuznetsova E.S. Detection of Parts Obtained by Turning Methods Using the YOLOv5 Object Recognition Model. Vestnik Magnitogorskogo Gosudarstvennogo Tekhnicheskogo Universiteta im. G.I. Nosova [Vestnik of Nosov Magnitogorsk State Technical University]. 2025, vol. 23, no. 3, pp. 122-128. https://doi.org/10.18503/1995-2732-2025-23-3-122-128
1. Suzdaleva N.N. The potential of using neural networks by industrial enterprises in the context of Russian reality. Regionalnaya i otraslevaya ekonomika [Regional and sectoral economics], 2022;(11(173)):91-94. (in Russ.)
2. Altunina K.A., Sokolova M.V. Application of neural networks for modeling the turning process. Vestnik TGTU [Bulletin of TSTU], 2016 22(1):122-133. (in Russ.)
3. Stepanov D.V., Makarov A.V., Molotov A.M., Bolotov E.N. Convolutional neural networks for detecting defects and damage to structures. Promyshlennoe i grazhdanskoe stroitelstvo [Industrial and civil engineering], 2024;(9):52-58. (in Russ.)
4. Veretelnikov A.S., Gavlitsky A.I. Application of artificial intelligence in the metalworking industry. Elektronniy nauchniy zhurnal «Dnevnik nauki» [Electronic scientific journal "Diary of Science"], 2022;(12(37)). (in Russ.)
5. Alkhanov A.A. Machine learning and its application in the modern world. Problemy nauki [Problems of Science], 2021;(7(66):25-27. (in Russ.)
6. Kuznetsov S.V., Rogovik A.A. The prospect of using neural networks to plan the loading of a machining site. [International Journal of Humanities and Natural Sciences], 2024;(9):115-117. (in Russ.)
7. Mitrofanov S.P. Nauchnaya organizatsiya mashi-nostroitelnogo proizvodstva [Scientific organization of machine-building production]. Leningrad: Mashi-nostroenie, 1976, 712 p. (in Russ.)
8. Kuznetsov S.V., Anosov M.S., Rogovik A.A., Murugov Yu.S. Determination of the unit time coefficients of “shaft” type parts based on their similarity. Nauchno-tekhnicheskiy vestnik Povolzhya [Scientific and Technical Bulletin of the Volga region], 2024;(6):72-75. (in Russ.)
9. Kuznetsov S.V., Rogovik A.A., Murugov Yu.S. Selection of the optimal grouping of the "shaft" type parts to increase seriality in production planning. Mashinostroenie: setevoy elektronniy nauchniy zhurnal [Mashinostroenie: online electronic scientific journal], 2023;(1):51-55. (in Russ.)
10. Bochkovsky A., Wang C. YOLOv5: Creating State-of-the-Art Object Detectors in Real Time. [S. l.]. ArXiv. 2021:214-219.
11. Bochkovsky A. YOLOv7: Trainable Boosted Data-Driven Layer for Real-Time Object Detection. [S. l.]. ArXiv. 2023:5147-5155.
12. Redmon J., Farhadi A. YOLO9000: Better, Faster, Stronger. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017:7263-7271.
13. Redman J., Shoji F., Farhadi A. YOLOv5 Training and Improving Object Detectors and Segmentation Models with One Click. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2022:962-972.
14. Lin T.-Y., Dollár P., Girshick R., He K., Hariharan B., Belongie S. Feature Pyramid Networks for Object Detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017:936-944.
15. OK 012-93. All-Russian classifier of products and design documents (ESCD classifier) Moscow: Standards Publishing House, 2004, 40 p. (In Russ.)