#10 2022

CONTENTS №10

Levin S.E., Bogdanova L.M., Senatorov M.Yu., Naghibin S.Ya., Kurpatov O.V.  Interval estimation of disaster cost in the safety pyramid of industrial site

The paper offers a method for a priori calculation of C1-C4 safety class boundaries dependent on economic losses from disasters and incidents in oil refining segment. The method allows for both direct and indirect losses. The interval estimates of industrial safety class boundaries can be used for a priori calculation of possible losses from industrial disasters and safety incidents.

Keywords: industrial safety incident, industrial safety pyramid, event cost, integral risk, remote monitoring.

Kudryashov V.S., Alekseev M.V., Ivanov A.V., Kozenko I.A., Tokareva O.A.  An automated control system for grain reception, cleaning, dehydration, and storage complex

The paper presents an automated control system for the elevator complex of Volovsky formula-feed plant JSC. The system features an algorithm for automatic development of grain transportation and dehydration routs as well as regulatory control algorithms for process parameters. The controller is programmed in TIA Portal environment. Controller settings in the temperature loop were calculated with the help of a program, which implemented a numerical gradient optimization subject to minimum integral square error.

Keywords: grain transportation and dehydration, PLC, automatic rout development algorithm, monitoring and regulatory control of process parameters, optimization, numerical gradient method.

Discussing a Topic…

Equipment monitoring and technical diagnosis in industry

Boyarkov D.A., Kazymov I.M., Kompaneets B.S.  Application of fuzzy logic tools in power network maintenance

Adequate assessment of electrical equipment state is becoming a task of a growing importance. The paper describes the application of fuzzy logic tools for estimating mechanical defects of electrical equipment. The algorithm provides an estimate allowing for heterogeneous quantitative and qualitative input information, partially and uncertain. Such approach enables the most objective maintenance and repair of power network thus work programs of power companies more efficient.

Keywords: power network, electrical equipment, maintenance and repair, fuzzy logic, membership function, uncertainty of information, mathematical model.

Antonenko I.N.  Comprehensive approach to reliability management of machinery and equipment

A key statement of GOST R 55.0.00 series (Russian National standard) asserts that asset management decisions should allow for the needs and expectations of all interested parties. In practice, however, the decisions are often made inside the organization without discussing with all stakeholders. This refers also to reliability management. Managerial decisions in case of equipment failures and unplanned downtimes are often superficial and boiled down to the punishment of those ones who are directly responsible for the reliability. The paper offers an approach, which ensures of stakeholders participation in reliability management.

Keywords: reliability management, maintenance and repair, total cost of asset ownership, overall equipment efficiency.

Maksimova O.V., Nikolaev P.V., Moyseenko S.V.  Automated monitoring systems for thin-film electroluminescent indicator parameters

The paper examines the parameters of thin-film electroluminescent indicators (TFEI) that should be monitored during fabrication and subsequent operation and maintenance. It offers a data acquisition algorithm including the measurement of optical and electric parameters as well as external factors. A measuring system's structure is presented as well as the TEFI monitoring software architecture.

Keywords: thin-film electroluminescent indicator, brightness, data acquisition, measurements, test program.

Discussing a Topic…

Application of neural networks in image recognition

Krasnov A.E., Golovkin M.E., Nikol'sky D.N., Blagoveshchensky V.G.  Wave network for image recognition

The technology of chromatographic, spectral, and video data mining is a key area of technical support in food industry. Video methods enable comprehensive monitoring of edible raw materials and foodstuff based on their image textures and forms. Deep learning technologies such are extensively used, in particular, convolutional neural networks. However, the description of images invariant to their rotations poses a challenge because a preferable image orientation is not possible in the analysis. This implies the need for hybrid convolutional network with sophisticated architecture. The paper examines the simple architecture of a new class of convolutional networks named wave networks. Based on coherent image blurring, they enable the development of their special descriptions or patterns, the recording of which for 23 rotations (0…45 degrees) ensures complete invariance to any image rotations. Coherent image blurring is based in sine and cosine Fresnel transforms, which allow for phase relationships in the relative position of its pixels that eliminates the degeneracy in the description of these positions. The paper shows that coherent blurring of images is preferable for the description of their various structural forms. Examples of binary and halftone image description forms are cited.

Keywords: wave network, image, coherent blurring, pattern, rotational invariance, form.

Malyshev G.S., Kamnev M.A., Drumov I.V., Terekhin G.V.  Application of variational autoencoder for extending training dataset of a neural network which recognizes the flaws on siliconized graphite articles

The paper outlines the theoretical background of variational autoencoders (VAE). It examines the opportunity and expediency of applying VAE for extending training datasets of neural networks for the cases where comprises of several tens of patterns. Examples of unpractical VAE application are cited; practical recommendations on the development of training datasets for VAE are made. The examples of images with the defects of images sampled from different parts of hidden factor distributions are discussed, an attempt of result explanation is made. The settings used in VAE encoder and decoder network training are provided. Aт image composition method based on VAE application is offered. The research results can be used for the detection and classification of flaws on siliconized graphite articles.

Keywords: neuron networks, encoder, decoder, variational autoencoders, hidden factor space, sampling, generative models, defects, minibatch, training dataset.

Ogorodnikov I.I., Musa Z.  Machine vision for ripe fruit identification by robot system

Machine vision is extensively used in automated and robotized production including agriculture and food industry for growing, gathering, and processing of fruit and vegetables. The paper presents a machine vision system for ripe fruit identification. An algorithm that identifies ripe fruit per their red color is developed along with another algorithm based on deep learning of a neural network, which is applicable for detecting any ripe fruit. Both algorithms are implemented in software code in Python language. An fruit detector is fabricated and tested, which detects a ripe fruit with the help of video camera and measures the object distance by an ultrasonic sensor. The instrument can be used together with a robot manipulator for gathering ripe fruit.

Keywords: machine vision, image processing, color detection, convolutional neural networks, deep learning, agricultural works.

Zhuravlev S.I., Sidak A.A., Kadomkin V.V., Matvienko V.A., Sidak D.A.  Regulatory and methodological aspects of security monitoring of automated information systems in industry

Fast development and dissemination of automated information systems in various industries are associated with new information security risks. To protect confidential information, personal data, information processed in process control systems of critical industrial facilities, and governmental information systems a variety of information security tools are used, which implement access control and information flows policies. At the same time, due to the complexity and the large scale of modern information systems and information infrastructure as a whole, the use of control tools and, above all, information security monitoring systems is gaining the importance. Currently, the regulatory framework for monitoring activities is being formed. Therefore, the development of regulatory and methodological aspects of security monitoring of information systems including the ones used in industrial automation, is of scientific and practical interest. These problems are overviewed and discussed in the article.

Keywords: automated information system in industry, information system, regulatory and methodological aspect, monitoring, incident, information security, security event.

Fedin F.O., Kodanev V.L.  Self-organizing maps ensure information integrity in automated systems of an enterprise

The paper describes the development and training of a neural network aimed at the segmentation of industrial enterprise's employees. The model allows to analyze the level of the potential hazard of employees (automated system's users) in order to ensure the integrity of the enterprise's automated systems. The segmentation of employees is done with the help of self-organizing maps.

Keywords: model, clustering, Kohonen maps, chart's node, learning sample, learning.

Osipova N.V.  An overview of methods and tools for measuring the load of ore grinding mills

The paper overviews commercial load level analyzers for ore grinding mills, which are used for automation and optimization of the grinding process as well as for overload protection. This include sound, vibration, magnetic, radiotelemeteric, electrical, radioisotope, mechanical, hydraulic, laser, and combined measuring techniques. Operation principles of popular domestic and foreign control instruments are analyzed, their merits and drawbacks are highlighted. Development outlook for a combined method for measuring mill loading by processing algorithm and data transmission is outlined.

Keywords: analyzer, ore grinding mill, mill load control methods.

Vedmid P.A.  Import replacement and the architecture of PLM/QMS systems

All foreign PLM/QMS systems feature monolithic architecture. The paper argues that import replacement reject obsolete software design concepts and use Open Source and microservices.

Keywords: PLM, QMS, monolithic architecture, import replacement, Open Source, microservices.