Smart steel: how neural networks help to evolve metallurgy

Why AI solutions are required in metallurgy?

Implementation of AI technologies in metallurgy enables to increase the production efficiency based on such criteria as:

• Speed – decrease in time required for performing technological and production tasks and for decision making.
• Quality – possibility to improve product and services features.
• Subjectivity – decrease in number of mistakes and errors related to the human factor.
• Personalisation – formulation of personalised offers and trajectories.

With the help from artificial intelligence the production chain can become more transparent and efficient – for example, such task as the scheduling of machinery operation. In such application, the combination of recommendation models with digital visualization of manufacturing layout, which in turn can increase profitability of manufacturing enterprise by 10%. Such level of operational improvement is often indicated by the research performed by the Alliance Research Group for industrial application of AI. Recommendation systems can also help with managing inventory via the control of crucial parameters in manufacturing: fluctuation in demand, profitability and losses due to storage not being fully optimised. This provides not only means for decreasing costs but also minimise the areas where such costs are arising in the first place.

One additional crucial criterion is the safety at the production site. Most commonly, occupational health and safety compliance is monitored by the operators of the production processes, but according to the research, they most often cannot cover supervision of safety at entire production plant. Solution for this problem is presented in form of AI model that processes all information live and formulates recommendations based on historic data. And, as an example, these produced recommendations are utilised by the production facility personnel to manage melting procedure.

Implementation of artificial intelligence also enables to optimise planning of maintenance and repair works. AI systems are already capable to take into consideration multiple of variables, such as constraints in the number of available personnel, financial and technical resources, production priorities. For instance, by collecting information about each defect live, the technological solution can update enterprise plan for maintenance and repair works for several weeks ahead of time. This is not only capable to decrease operational expenses but also makes it more convenient for planning engineers to perform they tasks, say the authors of the research.

What solutions are already implemented and in-use?

Mallenom Systems IT company in partnership with Data-Center Automatika has developed a system for in-plant logistics for NLMK (Novolipetsk Metallurgical Plant), which is essentially serves a role as a digital twin of steel production plant. Algorithms for video analytics developed for this system enable to collect actual data for the location and condition of cranes, crucibles and crucible transportation mechanisms in two facilities at the production site. Based on this data the created system performs planning and operational re-optimisation for the utilisation of these factors of productions. Based on the analysis performed by the client, the estimated economic impact from AI system implementation will be equivalent to 1 million of Euro per year in reduction of operational expenses. This accounts for the decrease of electricity expenses, and more efficient use of graphite electrodes and aluminum wire rods. The reduction of operational expenses and more efficient use of factor inputs is achieved by decreasing wait time of the metal in a crucible for up to 7%. An intermediatory result which has been reached during first two month of the system test implementation stage was a reduction of the average operational expenses by 5%.

Additional system that was developed by Mallenom Systems has been a solution for reading and verification of the surface identification markings for metal pipes of various diameters. This system enables to detect pipes with identification marking that are either incorrect or have a poor quality of application at the earliest stage. Implementation of the system also provides the means for tracking of manufactured items across all production stages and as a result decreases the overall number of defects during manufacturing.

Finally, Mallenom Systems has developed a system for Severstal to control positioning of hot-rolled coils when they are being transported by a chain conveyor. AI-based solution detects dangerous offsets of the coils which may lead to their fall from the conveyor line. As a result, the system prevents potential industrial accidents, damage of manufactured items and disruptions in the collection of data by track and trace system.