RSS
Translated by
Project

As machine learning helps steelmakers. Case of NLMK and Jet Infosystems

Customers: Novolipetsk Metallurgical Combine, NLMK

Novolipetsk; Metallurgical industry

Contractors: Jet Infosystems
Product: Artificial intelligence (AI, Artificial intelligence, AI)

Project date: 2019/09  - 2020/03

Content

IT solutions based on artificial intelligence actively get on different sections of industrial productions. For example, on Novolipetsk Metallurgical Combine (NLMK) AI workplace – near the steelmaker who performs steel smelting.

Specialists of Jet Infosystems company were connected to a problem of transformation of artificial intelligence into the steelmaker. They in literal sense trained program system – it works at a basis of algorithms of machine learning.

Why to artificial intelligence to understand steel grades

For receiving steel grades with the improved characteristics, for example, intended for specific conditions of operation during smelting enter additional chemical elements into its structure. The most widespread method is introduction to liquid fusion of metal of special materials in the form of iron alloy with one or several chemical elements (silicon, manganese, chrome, etc.). They are entered into the different periods of smelting and processing of steel. For example, ferro-nickel is entered into the first period – oxidizing because nickel is not oxidized in the oxygen converter, but contains hydrogen which when heating turns into gas and is removed at the second stage in the course of steel boiling. And here ferroniobiya and ferrovanadium well are oxidized therefore they are entered at the final stages of processing of steel.

Steel grades differ among themselves with chemical composition for which in each case there is accurate "compounding". However the processes happening when smelting and processing became, are so difficult that adding of a certain amount of ferroalloy does not guarantee exact hit in an interval of admissible chemical parameters. There are many reasons for which it is impossible to get from the first to "gate" of admissions, for example, there can be too much oxygen, etc. This process is similar to art, only instead of creative inspiration experience a little here. If the experienced specialist who meticulously studied all smallest parts of processes works at this section of metallurgical production, he considers parameters of melting more precisely and quicker receives the necessary result, than the beginner who only masters a profession – it should undergo several iterations, adding ferroalloy and carrying out chemical analysis of the turned-out metal.

Converter workshop NLMK

«
The task which was set by plant is clear: define minimum necessary amount of ferroalloy to get to the set interval on chemical composition, to use at the same time the minimum quantity of materials, and if a possibility of the choice, then to apply the cheapest material, – Evgeny Kolesnikov, the director of the Center of machine learning of Jet Infosystems company tells.
»

As it is possible to optimize steel melting process

«
In terms of digital solutions, it is classical decision making support system which is designed to bring up all workers to the uniform level of skill and at the same time to optimize production processes, – Evgeny Kolesnikov explains. – Of course, at any enterprise there are true talents, venerable professionals – they are capable to work more precisely than a computer system. But on average the machine algorithm proposes more optimal solutions as the person is inclined to be reinsured.
»

The solution of optimization of melting at the first stage of the project was tested on the platform of the first converter workshop NLMK on a limited range of steel. The solution includes two key components:

System of the forecast of chemical composition. System "heart" – a mathematical model based on algorithms of machine learning (ML) which predicts what will be chemical composition if to specific timepoint to add these or those materials.

«
It allowed to create model which with a big accuracy works in the forecast mode "that if, – Evgeny Kolesnikov tells.
»

To the steelmaker of NLMK smart computer programs help to work

«
In terms of mathematical statistics, it allows to reach higher accuracy of the forecast, – the expert continues.
»

Using this digital instrument it is possible to perform quickly search of a large amount of different ferroalloys and to select those which give the greatest economy from these combinations.

«
At the first stage of the project we conducted an in-depth study of production processes", – Evgeny Kolesnikov says. – Learned target chemical composition of steel and ferroalloys, their interchangeability, a melting route. Obtained data on the cost of each ferroalloy, and a remaining balance of materials in a warehouse. On the basis of all this set of data we taught to select a system such combination of components of ferroalloy which has the smallest cost, but at the same time gives required chemical composition.
»

Technical implementation

Recommendatory service is implemented as the automated workplace of the steelmaker. Recommendations are output to the monitors located in premises of control of different units of workshop. Within the project a new system was integrated with external data sources, including the systems of SAP and an APCS.

Crucial element of integration mechanisms – communication with uniform corporate storage of "crude" data Data Lake of all plant — all historical data of NLMK are stored in it.

Besides, last year at the enterprise the analysis system of data and modeling was brought into productive operation (we SIT DOWN) own development. We SIT DOWN in a basis of an architectural concept the structure of Data Lake which allows to obtain data in a "crude" type is put and to save them for further processing by the easiest and economic way.

The management of smelting of steel on NLMK reminds work of the situational center

«
The IT infrastructure which is already created on NLMK allowed to implement an implementation project of recommendatory service as normal service of system integration, no particular problems arose, – Evgeny Kolesnikov emphasizes.
»


In 2019 the system working in converter workshop No. 1 is transferred to the mode of commercial operation. NLMK already started development of similar recommendatory service on the second platform – in converter workshop No. 2. By plant's management estimates, the expected economic effect at expansion of service on the maximum volume of a branded range in two workshops of converter production can be 100 million rubles a year.

The Jet Infosystems company says that the service assistant developed for NLMK in optimization of swimming trunks has very good perspectives of replication on other enterprises of ferrous metallurgy of the Russian Federation. According to the experts the companies, in the country about 100 metallurgical companies use ferroalloys in production.

«
Physical and chemical processes of melting do not depend neither on specific production regulations, nor on pattern of ownership, nor on the size of the plant, – Evgeny Kolesnikov emphasizes. – Even small steelmaking production is billion turnovers, and economy on ferroalloys for it is very notable. It offers good prospects of implementations of such IT solution, as on domestic enterprises of ferrous metallurgy, and abroad.
»