#3 2026

CONTENTS №3

Discussing a Topic…

Modern solutions for automation and digitalization of A-plants

Grigoriev M.S., Maslova K.A.  Analyzing the compliance of HMI of personnel workstations at the main control room of Leningrad A-Plant #2 with present-day design trends

Basic HMI design concepts are examined with the example of the project of the main control center of Leningrad A-Plant #2. HMI types and the approaches to information representation are described. The developed interface is assessed for compliance with the principles defining situational awareness of operators, as well as modern approaches to control and information support for operational personnel.

Keywords: workstation, main control room, video frame, human-machine interface.

Zharko E.F., Abdulova E.A., Chernyshev K.R.  Hybrid intelligent system for A-plant operator support. Architecture and verification methodology

Architectural and methodological solutions are proposed for the verification of a hybrid intelligent support system for A-plant operators. The system combines machine learning, expert rule base, and a digital twin. The methodology is built as a multi-level verification and validation process, comprising component-by-component verification (including static constraints, training reproducibility, and dynamic testing of machine learning (ML) models), formal verification of logic and rule consistency using SAT/SMT approaches, as well as integration verification of thr consistency of outputs, time constraints, and controlled degradation modes. To identify boundary cases, it is proposed to use metamorphic testing of ML components based on physical invariants and operational constraints, as well as scenario-based checks of explainability and traceability of recommendations. The paper shows that the proposed approach makes it possible to identify inconsistencies between heterogeneous subsystems and reduce the risk of hidden defects before the system validation stage.

Keywords: hybrid operator support systems, verification, validation, A-plant, machine learning.

Baibulatov A.A., Poletykin A.G., Byvaikov M.E.  Upper-level control software for A-plants: automation of changes

Application software is a key to successful implementation of all upper-level control functionality of an A-Plant. It features increased complexity, long-term maintenance period and the need for multiple changes within the control system’s lifecycle. The paper presents the solutions developed and implemented in the Institute of Control Sciences of the Russian Academy of Sciences for the automation of changes in application software. It outlines the automation object, describes the change making task and proposes key ideas of the solution. The specialized tool for automated changes is the data preparation system. Its structure, hard- and software tools are described.

Keywords: application software, maintenance, data preparation system, upper-level control system, process control system, A-plant.

Kalashnikov A.A.  The design of automatic calibration systems for A-plant measuring channels

Since 2020, research and development work has been initiated in the nuclear energy industry to automate calibration procedures for measuring channels of process control systems. Between 2021 and 2024, the developed automatic calibration methods and technologies were successfully tested in practice: bench tests had been conducted on pressure measurement channels with DM5017 and AIR-30M sensors. As a result, the technical feasibility and advantages of the developed automatic calibration methods were confirmed in relation to the current level of technological development, including foreign control systems from Yokogawa and Emerson. The paper further develops the topic and discusses the approaches and design concepts established in 2025-2026 and exemplified the first pilot system for automatic calibration of measuring channels at an A-plant.

Keywords: automatic calibration, measuring channel, process control system, A-plant.

Faruntsev S.D., Shkodyrev V.P., Gebel E.S.  Multi-level supervisory optimization of process mode at oil treatment units

A hierarchical approach to solving the problem of supervisory optimization of a multilevel object was developed and exemplified in the model of oil treatment and pumping facility (OTPF). The OTPF includes two parallel oil treatment units with the sections of preliminary and deep dehydration of oil emulsion. The paper formulates the general criterion for the whole facility and the local ones for its individual sections in the context of variable technological and market situations. It describes analytical and experimental control models for key process equipment of the facilities such as preheaters, flash drums, precipitators, and the unit as a whole. A method was used to form local compromise (Pareto efficient) control sets for operating modes of the key equipment pieces with the subsequent approximation of those sets using neural network models. A multi-level optimization algorithm is proposed and implemented on the OTPF model, which searches for the optimal mode in the coordinate space on local Pareto sets. This significantly reduces the problem’s dimension, ensures convergence and local efficiency of the operating modes of each equipment pieces, and facilitates adaptation to changing technological and economic conditions.

Keywords: hierarchical control object, oil treatment and pumping facility, multi-level supervisory optimization, analytical and experimental models, optimization on local Pareto tradeoff sets.

Volkov E.A., Smyslova A.L.  Development and implementation of predictive model for thermal conditions of a blast furnace

The paper examines the problem of predicting thermal conditions of a blast furnace as a complex inertial object with distributed parameters in the absence of direct measurements of internal states. It proposes an operational metric of heat input into the bottom zone. It allows for the contributions from hot blast, and the combustion of coke and natural gas with reference to the heat content of the outgoing tuyere gas. The metric is used as a physically interpretable indicator of the thermal state and an input feature for forecasting models. Based on the moving median and standard deviation of heat input, a statistical-adaptive self-regulation mechanism has been implemented, which discerns "green", "yellow" and "red" zones thus making it possible to generate the recommendations for adjusting fuel and gas blower modes under persistent deviations. Industrial application results showing the potential for reducing variability in iron silicon content and coke consumption are discussed, as well as prospects for further development of hybrid models combining physical-chemical equations with machine learning methods for intelligent decision support and predictive control of blast smelters.

Keywords: blast furnace, thermal conditions, bottom zone of the furnace, heat input, silicon content in the cast iron, statistical process control, moving median, standard deviation, control zones, predictive control.

Ivanov V.A., Denisov L.A., Shchapova L.V.  Research and selection of optimal operating modes for a group of technological sections of an oil pipeline network using a genetic algorithm

The paper discusses the selection of optimal operating modes for a group of adjacent technological sections of an oil pipeline network. The criteria of minimum specific energy consumption and efficiency of a process section can be used when implementing a plan for delivering oil to the consumer. To improve the efficiency of adjacent process sections, it is proposed to reduce the number of transitions by minimizing the product imbalance in the tank farm. A model of a group of oil pipeline sections has been developed for studying operating modes. Calculation results are included, which enable the development of recommendations for the operation of an oil pipeline network in modes featuring low specific energy consumption coupled with high rates of equipment resource utilization with a minimum number of switching operations.

Keywords: oil-trunk pipeline, operating mode, mathematical modeling, genetic algorithm, optimization.

Sazontiev R.I., Susarev S.V.  A system for monitoring and forecasting the rate of growth of corrosion damage in oil and petroleum product tanks

The problem of ensuring the reliability of vertical cylindrical corrodible steel tanks is discussed. The relevance of creating an auxiliary tool for improving the reliability of tank and tank farm operation is confirmed. An intelligent system for monitoring and predicting the rate of corrosion of tank walls was developed based on a machine learning algorithm. Its development and performance details will be discussed in the next article. The model allows for the parameters such as product temperature, steel grade, and the type of petroleum product stored, and enables the calculation of wall thickness dynamics over the service life. The predicted corrosion rate is integrated into the stress-strain state calculation of the structure, which enables the assessment of the remaining service life and timely decision-making on tank maintenance. The results of numerical modeling are presented for two operating scenarios: without changing conditions and with the switch to stock-tank oil storage. It is shown that the proposed approach makes it possible to increase the accuracy of forecasting and ensure the safe tankage operation through early detection of critical conditions.

Keywords: monitoring, corrosion, vertical cylindrical steel vessel, machine learning, residual life, forecasting.

Bazenkov N.I., Efremov A.Yu., Zhilyakova L.Yu., Kuznetsov O.P., Kulinich A.A., Maksimov D.Yu., Sukhoverov V.S. Application of network models in social, technical, and biological systems. Part 1. Knowledge representation models

The paper presents the first part of an overview of network models that are used for controlling complex systems of various natures. It describes knowledge representation models such as weighted and symbolic cognitive maps and their modifications, formal ontologies describing hierarchical structures of subject areas. Classical graph models and distributed algorithms are presented that make it possible to restore the global knowledge about the structure of the graph as a whole from local data known to each of its vertices. For each of the described models, application examples from various subject areas are given.

Keywords: cognitive maps, fuzzy cognitive maps, formal ontologies, distributed calculations on graphs.