The paper presents a methodology for training neural networks for vision tasks on synthe-sized data on the example of steel defect recognition in automated production control systems. The article describes the process of dataset procedural generation of steel slab defects with a symmetrical distribution. The results of training two neural networks Unet and Xception on a generated data grid and testing them on real data are presented. The performance of these neural networks was assessed using real data from the Severstal: Steel Defect Detection set. In both cases, the neural networks showed good results in the classification and segmentation of surface defects of steel workpieces in the image. Dice score on synthetic data reaches 0.62, and accuracy—0.81.

Big Data analysis is becoming an everyday task for companies all over the world, including Russian companies. Due to advances in technology and the reduction in the cost of storage systems, companies can now collect and store large volumes of heterogeneous data. The important step of extracting knowledge and value from such data is a challenge that will eventually be met by all companies seeking to maintain their competitiveness and place in the market. However, companies face several challenges when it comes to collecting, pre-processing, and integrating data into cohesive data sets designed to deliver analytics. In this article, the above problems and possible solutions are illustrated using the example of cleaning, integration, and normalization of data obtained in the measurement of indicators for the Vanyukov melting furnace process. The article considers an approach to the study of metallurgical processes using the analysis of large operational control data sets. Standard methods of processing the data sets of operational process control are used. The correlation analysis of the main process parameters is carried out. The results are interpreted for their further practical application.

Qualification of employees who operate technological processes directly influences the safety of production. However, the employees’’ qualification cannot completely exclude human factor. Today, there are many technologies that can minimize or eliminate human factor impact on production safety ensuring. The augmented reality technology is an example of this technology. Nowadays, the augmented reality technologies and industrial technologies integration process moves to a new level of development. These technologies have huge experience, which has been accumulated in a long period of time. -This new level turns available by this experience combination and integration; it brings additional profit to the enterprise and can be a basis for completely new technologies. This paper shows an example of combination of augmented reality technology and oil pumps maintenance. For researching of efficiency of augmented reality system for oil pump maintenance, the laboratory unit with Grundfos vertical electric centrifugal pump (CR15-4 A-FGJ-AE-HQQE) was used. The laboratory unit is a physical model of one of the continuous oil processes. The oil pump of this laboratory unit is object of this research. The algorithm of servicing of oil pump was developed. The test of system and algorithms were carried out with four groups of people: the first one had only instructions to use on hand, the second one only used the internal recommendations of the system, the third one used only the help of an expert, and the fourth used internal recommendations and, if necessary, contacted the expert. The results show the efficiency and actuality of augmented reality technology for maintenance of industrial equipment, especially for the equipment operated in remote Arctic conditions.