Artificial Intelligence in Manufacturing

The main articles are:

• Artificial intelligence (AI, Artificial intelligence, AI)
• Information Technology in Industry

2024

How Russian enterprises are introducing machine vision

Russian industrial enterprises are actively mastering machine vision technologies, integrating them with artificial intelligence to increase production efficiency. The growing interest in this area became known in October 2024 from the results of a study conducted by the Institute for Statistical Research and Knowledge Economics of the Higher School of Economics. Read more here

80% of AI solutions used in Russian production are created in the Russian Federation

At the end of 2023, approximately 80.9% of AI solutions used in Russian production were created in the Russian Federation or significantly modified by domestic developers. For comparison, in 2020, the indicator slightly exceeded 73%. Such data are given in the review of the Institute for Statistical Research and Economics of Knowledge of the Higher School of Economics, published on July 1, 2024.

It is noted that a narrow circle of organizations is engaged in the creation of advanced production technologies of artificial intelligence (PPT AI) in Russia: from 2020 to 2023, their number increased from 35 to 74, which is about 7.5% of the total number of PPT developers. At the same time, the intensity of using AI systems in the Russian Federation is increasing. Moreover, the use of technologies created by organizations independently is growing at the highest pace: in three years (by the end of 2023), their number has almost tripled.

source = HSE

During 2023, 88 AI PPTs were developed in Russia, which is approximately 3.2% of the total number of PPTs created in the country. Ten of these products are fundamentally new systems that have no global analogues. In the total volume of new AI solutions, 31 products were developed by organizations operating in the field of information and communications. Another 26 solutions were released by developers from the field of higher education, 19 from the scientific field.

The review also says that in Russia the practice of using AI PPTs as an independent technology in the production of products is limited: in 2023, only 634 organizations applied such solutions in their activities. Of these, 372 enterprises (almost 60%) work in the field of information and communications. The number of AI BSA used since 2020 has increased 1.8 times (from 582 to 1030), but their specific weight in the total number of BSA used in production does not exceed 0.5%. And more than 60% of AI solutions used were implemented by organizations during 2021-2023^[1]

2023

AI in heavy industry: prospects and directions of use in Russia

According to the annual report, in Center for the Development of Artificial Intelligence under the Government of the Russian federation 2023 artificial intelligence , 25% of companies operating in used technology. industries About 30% of organizations announced their intentions to use these technologies in the next three years. artificial intelligence Egor Sachko, extractive industries an expert on artificial intelligence, who implemented large projects of operational transformation using AI for a number of companies in heavy industry in and abroad, spoke Russia about how production management changes, and what trends and difficulties exist in this process. More. here

Created an open database for designing materials with specified properties using AI

On May 10, 2023, Innopolis University reported that a team of researchers from Russia and Singapore formed an open database for designing materials with specified properties using artificial intelligence tools.

The project was attended by the rector and employees of Innopolis University, experts from the National University of Singapore and the Higher School of Economics, as well as Nobel Prize winner in physics Konstantin Novosyolov. The published library of several thousand two-dimensional materials contains information about the structure and properties of single-layer materials with point defects.

Scientists from Russia and Singapore have created a database for designing materials with specified properties using AI

Scientists note that the development of solar panels, photocatalysts and biochemical sensors requires two-dimensional materials designed with the addition of impurities and defects. However, such compounds are difficult to find using classical calculation methods using quantum chemistry. The solution to the problem, according to the authors of the work, may be the introduction of machine learning tools.

To get a material with certain properties, you need to know the relationship between the structure and the property of the defects that need to be added. This is a difficult task, given the huge number of possible starting materials and configurations of defects. Machine learning methods make it possible to speed up the study of materials, namely, to reduce the number of experiments hundreds of times and generate the necessary structures for given properties, "said Ruslan Lukin, head of the Laboratory of Artificial Intelligence in New Materials at Innopolis University.

In the developed database, the datasets will be divided into two groups: defects with low and high densities. The information array mainly presents replacement defects, vacancies and their combinations. The published library includes approximately 3,000 calculated materials and 7,000 high-density defects. In the future, it is planned to develop machine learning models for more accurate and effective forecasting of material properties.^[2]

2020: Siemens: AI-based decisions will make key decisions and help make production safe

Over half of industry leaders believe that over the next five years, the world will transfer management of assets of great value to solutions based on artificial intelligence - in particular, factories, equipment and machines. This global trend was identified in a joint study between Siemens and Longitude Research. More than 500 top managers from the energy, production, infrastructure, transport sectors, as well as from the heavy industry sector took part in the survey on the development and implementation of AI, Siemens reported on October 26, 2020.

During the study, respondents were asked the following questions: what if you could automate a number of everyday operating solutions in your organization so that employees could focus on strategic projects such as developing new product lines or expanding your business? How good should the AI model be before you're ready to hand it control? Should its performance be at the level of engineers, or should it exceed it? What if a mistake can lead to serious financial losses or even injuries? These and other scenarios were proposed to 515 top managers of the industrial sector (including in the fields of power, production, heavy industry, infrastructure and transport).

The study showed that the level of confidence in AI is already very high for 2020:56% of respondents prefer to implement an ideal AI model instead of finding an experienced employee (44%). That means the other 44% likely have more confidence in decisions made by people, even if the evidence suggests in favor of AI.

In addition, the study pays attention to the types of contextual data that, according to leaders, can be considered the most useful at the time of the survey. Most of the votes (71%) on the issue of the most important and insignificant advantages were given by the participants in favor of data from equipment manufacturers. They are followed by internal data from other departments, regions or departments (70%), supplier data (70%) and performance indicators of products sold when used by customers (68%).

In these industries, many use cases assume the possibility of using AI in order to avoid accidents and make workplaces safer. In this regard, it is worth noting that, according to 44% of respondents, over the next five years, AI-based systems will autonomously monitor machine equipment, the operation of which carries potential risks of injury or death of personnel. Even more respondents - 54% - believe that in the same time frame AI will autonomously control individual assets of great value of their companies. But in order to transfer such responsibility to industrial AI, it, according to the survey participants, must reach the next level. In most cases, new approaches to data management, collection, mapping and sharing will contribute to this.

These include, for example, knowledge graphs that reflect the relationships between objects and connections in different data sets, or digital twins that allow you to create detailed digital models and simulate the behavior of real systems, assets or processes. Using industrial knowledge graphs to improve AI models by combining different datasets has very high potential.

"Knowledge graphs add context to the data being analyzed," said Norbert Gaus, head of scientific research in the field of digitalization and automation at Siemens. - For example, the technical characteristics of the machine can be analyzed in the context of design data, including the tasks for which the machine is intended, the temperatures at which it should operate, key parameters built into components, etc. Add to this the service history of similar machines, including malfunctions, reviews and expected results of inspections over the entire life of such a machine. Knowledge graphs make it much easier to link the industrial data needed to train AI models and add valuable contextual information. "

2017: AI helps Carlsberg create new beers

In December 2017, Carlsberg announced the use of artificial intelligence, which helps the Danish company create new beers. Read more here.

Robotics