The Oskol Electrometallurgical Plant (part of the Metalloinvest company) is completing pilot operation of an automated system for recognizing the stamp on cast billets in section rolling shop No. 1. The idea of the project came from the OEMK employees during the innovation competition organized by the JSA Group (part of the multidisciplinary IT group "IKS Holding") at the enterprise.
Our own intelligent solution based on neural network technologies was developed by STI scientists NUST MISIS together with the plant's specialists. Software engineers from Metallo-Tech LLC were responsible for integrating the software part of the system into production processes.
“Previously, the operator manually checked the stamp numbers with the list in the system, selected a furnace for melting and then sent the workpiece to it, - says Evgeny Tkach, head of the section of section rolling shop No. 1 of OEMK. - The employee had no more than 40 seconds to complete these tasks, which created a large workload and increased the likelihood of errors. The introduction of automatic recognition has significantly reduced these risks. ”
The implemented solution uses five neural networks, each of which performs its own separate task. In the process of collecting data for their training, more than 60 thousand pictures of the ends of blanks with a stamp were taken.
"One version of the stamp on the blank is printed by a machine, the other, if the number is corrected, by people using paint," explains Dmitry Poleschenko, Associate Professor at the Department of Automated and Information Management Systems of STI, NUST MISIS. - At first glance, the recognition process is similar to reading numbers on cars. But when analyzed, such systems did not give the desired result. ”
“ This system was the first product developed as part of the Company's Digital Transformation program. A special feeling of pride is the fact that this solution, together with colleagues from MISiS, was created by our internal team, - said Yulia Shutkina, Director of Digital Transformation of Management Company Metalloinvest. “We will continue our course towards the development of Russian technologies and the application of best practices in production.”
The experience gained during the implementation of this project is planned to be used to solve similar problems. For example, video analytics technology can be implemented in determining the granularity of the composition of incoming raw materials on a conveyor, to determine the quality of an excavator bucket tooth, and even to track the use of personal protective equipment by enterprise employees.
