Development and Industrial Operation of an Intelligent System to Detect Markings of a Cast Billet at A. A. Ugarov Oskol Electrometallurgical Plant.

This article discusses the developmental procedure of a system that detects the markings of cast billets at the continuous furnace metal loading desk of the rolling shop No. 1 of the A. A. Ugarov Oskol Electrometallurgical Plant and describes its operation results. This system is based on neural networks that recognize the digits of a marking code. This paper proposes the recognition system structure as a whole, selects the structure of its constituent neural networks, and determines their tasks. The recognition results of the marking digits and form in which they are visualized to the operator at the control desk are presented. Based on the results, the accuracy of the industrial operation system for a continuous flow of billets is approximately 90%, taking into account possible distortions and heterogeneous structure of marking digits. The implementation of such a system enabled to increase the traceability of billets within the metallurgical plant and to reduce the operator load at the loading control desk as the developed system assists in the selection of billets for confirmation. [ABSTRACT FROM AUTHOR]