Severstal’s Karelsky Okatysh Cuts Sintering Emissions using ML

Russian steel giant Severstal’s iron ore mining and processing plant Karelsky Okatysh has tested a digital model for controlling the roasting machine. It helps to save fuel oil and reduce emissions of carbon and sulfur oxides during pellet burning, while maintaining the high quality of the finished product. The solution was developed by the specialists of Severstal Digital, together with the employees of the concentrate and pellet production department and the technological automation and metrology department of Karelsky Okatysh. Large amounts of data from Data Lake and the plant's archives were used to train the model. The system analyses the parameters of the raw pellets and predicts what the cold strength will be after firing. The information goes to the operator's screen. If the planned quality of the pellets is predicted, the model controls the calcining machine independently, ensuring the optimal operation of the burners. At the same time, it takes into account the current parameters of the operation of other machine units. In the event of an emergency or deviation in the predicted quality of the finished product, the operator turns off the digital assistant and controls the processes manually. Experts launched the control model on the roasting machine No 1 for the production of non-fluxed pellets. The pilot ran from December 2020 to March 2021. The specific consumption of fuel oil during the operation of the model decreased by 6.4%, which means a reduction in emissions of carbon and sulfur oxides. Now there is an active replication of the digital solution. A model for controlling the firing of fluxed pellets is already ready, which will be tested on the roasting machine No 1. In the development of a control model for the firing of non-fluxed pellets on the roasting machine No 3. Karelsky Okatysh is actively introducing digital technologies, and for these purposes in 2021 we plan to allocate 215 million rubles.

Severstal’s Karelsky Okatysh Cuts Sintering Emissions using ML

Related Stories