and made it possible to evaluate decisions
Posted: Sun Dec 22, 2024 9:37 am
Severstal
Severstal has been developing machine learning solutions for over six years. It all started with simple recommendation services that offered more optimal operating modes and looked for new opportunities. Later, advisors and analytical systems appeared. They began to predict quality, look for data relationships, made.
As technologies developed and improved, more complex structures using AI emerged. In particular, a whole korea phone number video inspection system was created that allows for the detection and classification of metal defects. It is represented by a line of products used at all key inspection points. They can record up to 40-60 classes of defects, which allows for the improvement of the quality of products supplied to customers. Previously, Western, more primitive technologies were used for these purposes.
"There are certification rules - conditionally, there should not be more than a certain number of defects per 1 m² of metal. The model can read thousands of defects - some scratch, corrosion - and help to certify the products. This system has entered the business process"
Potapova Svetlana
Potapova Svetlana
CEO of Severstal Digital
Severstal Digital CEO Svetlana Potapova said that thanks to machine learning (M.L.Machine learning (ML) is a branch of artificial intelligence that focuses on creating systems without explicit programming that learn and evolve based on the data they receive.) the company increased the productivity of the Stan 5000 line by 5.2% and ANGTs-4 by 3.4%, and also reduced the specific energy consumption by 2% due to the management of the compressor fleet. Computer vision (CVComputer Vision (CV) is a subfield of artificial intelligence that uses machine learning and deep learning algorithms to recognize and interpret objects in images and videos.) resulted in a reduction in emergency production downtime and reduced cases of employee injuries, while the quality of manufactured products only improved.
Currently, Severstal is working in four key areas of work with AI:digital twinsA digital twin (Digital Twin of Organization, DTO) is a software analogue of a physical device/object/product that models the internal processes, technical characteristics and behavior of a real object under the influence of interference and the environment., complex systems (ML+CV), video analytics platforms and digital assistants. Severstal also has the Steel View platform. It helps to promptly manage production, record technology violations and control the presence of personnel at workplaces.
NLMK
NLMK uses machine learning technologies to analyze large volumes of data, control product quality, and predict potential risks. Back in 2020, the Novolipetsk Iron and Steel Works developed and implemented a digital service that helps accurately dose ferroalloys during steel smelting. Artificial intelligence also monitors the tightness of coke oven battery doors in production. The IT system detects door depressurization within five seconds and alerts specialists. Machine vision and a neural network determine emissions in snow, rain, and when moving in the control zone of mechanisms. As a result, the problem is eliminated three times faster.
In April, the company announced tests of using AI to speed up the development of software solutions. They used several generative artificial intelligence tools (GenAIGenerative AI (Generative Artificial Intelligence, GenAI) is an artificial intelligence that generates new content in response to human prompts and instructions. It can create text, images, music, voice, and even video, imitating human creations with varying levels of quality.) to help developers write code. As a result of the pilot, certain types of work increased by 53%, and the execution of change requests accelerated by 34%. AI also helps technical support specialists categorize incoming emails to speed up communication and reduce the workload of specialists.
NLMK plans that in the future the model will be something like a chat bot: it will itself clarify information with users and respond to typical requests. The company is currently developing an AI assistant to create new training courses for the Corporate University, and is gradually introducingLLM modelsLarge Language Models (LLM) are neural network models that use machine learning algorithms to generalize, predict, and generate human languages based on large sets of text data.to assist in the selection of personnel.
Norilsk Nickel
Another giant of the Russian metallurgical industry that actively uses the opportunities offered by neural networks is Norilsk Nickel. Today, according to Alexey Testin, head of the company's digital technology development center, AI-based solutions have been implemented in almost all production areas.
"We have projects deployed in all divisions of the company from mining to metallurgical processing. Most of the solutions based on ML (machine learning) work ascopilotCopilot is a neural network assistant for programmers that analyzes millions of lines of code from open sources, generates suggestions and code fragments, offers hints and autocompletes while writing code. It can suggest entire functions or lines of code by analyzing the context of what has already been written., i.e. it controls the units in automatic mode. There are solutions, mainly for metallurgy, that operate in the "prompter" mode: the solution suggests adjusting the process mode, but does not make changes on its own"
Alexey Testin
Alexey Testin
Director of the Digital Technologies Development Center of Norilsk Nickel
Artificial intelligence recognizes the absence of necessary protective equipment on employees, thereby increasing safety at the facility. Recently, video analytics of underground equipment operations was introduced at one of Norilsk Nickel's mines. In the future, this technology is to be used at other mines.
Artificial intelligence not only monitors industrial safety at factories and controls product quality, but also helps manage the main technological processes in the company. For example, an assistant for the tax department based on a generative neural network allows employees to find the necessary document in a few minutes, and the Nika chatbot provides round-the-clock quick access to corporate services and information and allows managers to remotely approve employee requests for business trips and advance reports. In total, artificial intelligence-based solutions have already brought the company about 1% EBITDAEBITDA (Earnings before interest, taxes, depreciation and amortization) is an analytical indicator equal to the amount of profit increased by the amount of expenses on interest payments, income taxes, depreciation and accrued depreciation..
In the next three years, Norilsk Nickel wants to develop generative and predictive AI in the entire production chain in order to automate the search for information in reports and simplify document flow; relieve technologists from routine tasks; increase the speed of all services, increase the efficiency of equipment maintenance, and much more. The company's specialists are already working on creating an assistant that will be able to quickly provide a response based on 2 thousand patents, journals, technical instructions, laboratory tests, and other documents.
Application of AI in BIM
The functionality of modern BIM/TIM systems also uses the capabilities of artificial intelligence. For example, recommendations for optimizing architecture, structures, network routing and technological/production lines, searching for geometric and parametric collisions between systems, sections, elements, etc., automatic analysis of design solutions (information model) for compliance with regulatory requirements and standards. Reports, rules and requirements are received by the design engineer in real time, which means he can work more efficiently, faster, freeing up time to make more complex and balanced decisions.
AI can automatically generate 3D models of buildings and structures based on drawings and existing project data, analyze different design solutions and materials, choosing the most effective and cost-effective ones, monitor the progress of construction work, identify deviations from the schedule and propose measures to eliminate them, and predict capital investment estimates for the project before construction begins. All this leads to minimization of errors, which are most often caused by the human factor. This, in turn, reduces the risks of delays and budget overruns.
Severstal has been developing machine learning solutions for over six years. It all started with simple recommendation services that offered more optimal operating modes and looked for new opportunities. Later, advisors and analytical systems appeared. They began to predict quality, look for data relationships, made.
As technologies developed and improved, more complex structures using AI emerged. In particular, a whole korea phone number video inspection system was created that allows for the detection and classification of metal defects. It is represented by a line of products used at all key inspection points. They can record up to 40-60 classes of defects, which allows for the improvement of the quality of products supplied to customers. Previously, Western, more primitive technologies were used for these purposes.
"There are certification rules - conditionally, there should not be more than a certain number of defects per 1 m² of metal. The model can read thousands of defects - some scratch, corrosion - and help to certify the products. This system has entered the business process"
Potapova Svetlana
Potapova Svetlana
CEO of Severstal Digital
Severstal Digital CEO Svetlana Potapova said that thanks to machine learning (M.L.Machine learning (ML) is a branch of artificial intelligence that focuses on creating systems without explicit programming that learn and evolve based on the data they receive.) the company increased the productivity of the Stan 5000 line by 5.2% and ANGTs-4 by 3.4%, and also reduced the specific energy consumption by 2% due to the management of the compressor fleet. Computer vision (CVComputer Vision (CV) is a subfield of artificial intelligence that uses machine learning and deep learning algorithms to recognize and interpret objects in images and videos.) resulted in a reduction in emergency production downtime and reduced cases of employee injuries, while the quality of manufactured products only improved.
Currently, Severstal is working in four key areas of work with AI:digital twinsA digital twin (Digital Twin of Organization, DTO) is a software analogue of a physical device/object/product that models the internal processes, technical characteristics and behavior of a real object under the influence of interference and the environment., complex systems (ML+CV), video analytics platforms and digital assistants. Severstal also has the Steel View platform. It helps to promptly manage production, record technology violations and control the presence of personnel at workplaces.
NLMK
NLMK uses machine learning technologies to analyze large volumes of data, control product quality, and predict potential risks. Back in 2020, the Novolipetsk Iron and Steel Works developed and implemented a digital service that helps accurately dose ferroalloys during steel smelting. Artificial intelligence also monitors the tightness of coke oven battery doors in production. The IT system detects door depressurization within five seconds and alerts specialists. Machine vision and a neural network determine emissions in snow, rain, and when moving in the control zone of mechanisms. As a result, the problem is eliminated three times faster.
In April, the company announced tests of using AI to speed up the development of software solutions. They used several generative artificial intelligence tools (GenAIGenerative AI (Generative Artificial Intelligence, GenAI) is an artificial intelligence that generates new content in response to human prompts and instructions. It can create text, images, music, voice, and even video, imitating human creations with varying levels of quality.) to help developers write code. As a result of the pilot, certain types of work increased by 53%, and the execution of change requests accelerated by 34%. AI also helps technical support specialists categorize incoming emails to speed up communication and reduce the workload of specialists.
NLMK plans that in the future the model will be something like a chat bot: it will itself clarify information with users and respond to typical requests. The company is currently developing an AI assistant to create new training courses for the Corporate University, and is gradually introducingLLM modelsLarge Language Models (LLM) are neural network models that use machine learning algorithms to generalize, predict, and generate human languages based on large sets of text data.to assist in the selection of personnel.
Norilsk Nickel
Another giant of the Russian metallurgical industry that actively uses the opportunities offered by neural networks is Norilsk Nickel. Today, according to Alexey Testin, head of the company's digital technology development center, AI-based solutions have been implemented in almost all production areas.
"We have projects deployed in all divisions of the company from mining to metallurgical processing. Most of the solutions based on ML (machine learning) work ascopilotCopilot is a neural network assistant for programmers that analyzes millions of lines of code from open sources, generates suggestions and code fragments, offers hints and autocompletes while writing code. It can suggest entire functions or lines of code by analyzing the context of what has already been written., i.e. it controls the units in automatic mode. There are solutions, mainly for metallurgy, that operate in the "prompter" mode: the solution suggests adjusting the process mode, but does not make changes on its own"
Alexey Testin
Alexey Testin
Director of the Digital Technologies Development Center of Norilsk Nickel
Artificial intelligence recognizes the absence of necessary protective equipment on employees, thereby increasing safety at the facility. Recently, video analytics of underground equipment operations was introduced at one of Norilsk Nickel's mines. In the future, this technology is to be used at other mines.
Artificial intelligence not only monitors industrial safety at factories and controls product quality, but also helps manage the main technological processes in the company. For example, an assistant for the tax department based on a generative neural network allows employees to find the necessary document in a few minutes, and the Nika chatbot provides round-the-clock quick access to corporate services and information and allows managers to remotely approve employee requests for business trips and advance reports. In total, artificial intelligence-based solutions have already brought the company about 1% EBITDAEBITDA (Earnings before interest, taxes, depreciation and amortization) is an analytical indicator equal to the amount of profit increased by the amount of expenses on interest payments, income taxes, depreciation and accrued depreciation..
In the next three years, Norilsk Nickel wants to develop generative and predictive AI in the entire production chain in order to automate the search for information in reports and simplify document flow; relieve technologists from routine tasks; increase the speed of all services, increase the efficiency of equipment maintenance, and much more. The company's specialists are already working on creating an assistant that will be able to quickly provide a response based on 2 thousand patents, journals, technical instructions, laboratory tests, and other documents.
Application of AI in BIM
The functionality of modern BIM/TIM systems also uses the capabilities of artificial intelligence. For example, recommendations for optimizing architecture, structures, network routing and technological/production lines, searching for geometric and parametric collisions between systems, sections, elements, etc., automatic analysis of design solutions (information model) for compliance with regulatory requirements and standards. Reports, rules and requirements are received by the design engineer in real time, which means he can work more efficiently, faster, freeing up time to make more complex and balanced decisions.
AI can automatically generate 3D models of buildings and structures based on drawings and existing project data, analyze different design solutions and materials, choosing the most effective and cost-effective ones, monitor the progress of construction work, identify deviations from the schedule and propose measures to eliminate them, and predict capital investment estimates for the project before construction begins. All this leads to minimization of errors, which are most often caused by the human factor. This, in turn, reduces the risks of delays and budget overruns.