Machine vision “is watching” the environmental conditions

Published: Tuesday, 31 August 2021 08:44

Specialists of R&D MGTU implement several projects aimed at improving the environment. The projects are based on machine vision and neural network technologies applied at industrial plants to decrease harmful air emissions.

R&D MGTU developed a gas leakage automatic recognition system and technology for coke-oven doors and gas exhaust ducts of top structures. The machine vision technology detects coke-oven gas leakages. The system is tested at the Magnitogorsk Iron and Steel Works. As part of the project, over 30 video cameras were installed to monitor and register online 24/7 gas leakages at coke-oven batteries No. 7 and 8, and gas leakages of gas exhaust ducts of top structures at coke-oven batteries No. 7, 8, 13 and 14. If there are any deviations from setpoints, software and hardware register such events in an interactive log and send relevant people an email message with an attached photo indicating time, the battery and oven number.

“Gas leakages were monitored only visually, and now cameras can “see” much more than people. Thus, the R&D MGTU specialists have developed a system and a technology based on a neural network to detect gas leakages automatically”, explained Dmitrii Chukin, Project Manager. “Information about the rate and frequency of gas leakages at coke-oven batteries is registered in the database to carry out its statistical processing, resulting in additional conclusions: a wear rate of the door, need for its repair or replacement, and performance of workers who are responsible for gas tightness of doors”.

Modern equipment used in coking processes contributes to a lower environmental impact by reducing gross emissions of air pollutants and a better process discipline of personnel and state of workplaces. MMK and R&D MGTU intend to apply a continuous automatic control in other departments of the coke and chemical by-products division.

Another project of R&D MGTU aimed at improving the environmental conditions is associated with fugitive air emissions at facilities of PJSC MMK.

“Cameras installed near stacks and enclosing structures of shops register emissions online. Information is processed by a neural network that determines their time, period and rate. According to data received from the neural network, a signal is sent to a mimic panel, where it is converted into a visual image. The system controls main production processes of the works: coke and chemical by-products, blast furnace and steelmaking divisions”, comments Dmitrii Chukin.

People in charge will receive a prompt electronic notification, if there are air emissions. Real-time information is transferred to managers and specialists of shops and to the environmental protection laboratory, where the situation is analyzed and personnel make decisions, including technological ones, to prevent such emissions.