IZVESTIYA SFedU. ENGINEERING SCIENCES

This article addresses the pressing issue of ensuring the information security of critical information infrastructure (CII) amidst the qualitative evolution of cyber threats and the massive shift of threat actors toward stealthy "Living-off-the-Land" (LotL) attacks. Because such attacks are executed using legitimate administration tools, classical indicators of compromise (IOCs) lose their effectiveness, and traditional monitoring systems generate an excessive number of false alarms, thereby provoking "alert fatigue". To solve this problem, the paper proposes a methodology for the contextual validation of indicators of behavior (IOB) in Industrial Control Systems (ICS) networks. The primary scientific result is the developed "reverse enrichment" algorithm, which utilizes the deterministic nature of the technological process as a strict a priori filter. The algorithm verifies every control action by predicting the next state of the system using a state-space mathematical model and checking its membership in a formalized set of safe values, $\Omega_{safe}$. Additionally, the organizational context, $S_{org}$, including shift schedules and maintenance windows, is taken into account. Unlike probabilistic machine learning approaches, this method provides a strict binary criterion for command admissibility. The effectiveness of the proposed approach is confirmed by simulation results based on the verified dataset of the SWaT (Secure Water Treatment) academic testbed. The implementation of physical and organizational filters achieved a 97.7% False Positive Reduction Rate (FPRR) and successfully detected 1,021 out of 1,035 injected stealthy destructive impacts. The average computational latency of the algorithm was 1.4 ms, which fully satisfies the stringent requirements of real-time systems. The proposed method does not require the instrumentation of legacy field equipment and ensures the precise attribution of cyber incidents based on the physical laws of the production cycle.

The article explores the methods of data preprocessing and merging of multi-modal data. The data collected from multi-sensor devices contains noise, anomalies, and includes sensor failures. The use of raw data inevitably leads to false patterns when using machine learning models in which this data is used. To improve the quality of data, it is necessary to use special methods for preliminary processing and merging. The aim of the work is to create a hybrid preprocessing and multisensory data fusion controller operating at the peripheral computing level. To achieve this, the tasks of removing the consequences of technical failures and noise from the data, time synchronization, scaling of features and combining duplicate sources, considering their dynamic weight, have been solved. The following methods were applied: inter-quartile scale to detect outliers, linear interpolation to replace them, high-pass filter to suppress noise, and weighed multisensory fusion with dynamic weight calculation. In addition, methods for processing visual, sensory, and acoustic data within a hybrid pipeline are presented that can be adapted for implementation at the peripheral computing level. The experimental verification of the developed pipeline was carried out on a simulated set of climatic data with artificially introduced noises and anomalies.
The proposed approach has demonstrated high efficiency of data recovery. For air temperature, the coefficient of determination (R2) increased from 0.9807 to 0.9964 with a slight change in the mean absolute percentage error (MAPE) from 10.11% to 10.48%. For relative humidity, the R2 metric increased from 0.9248 to 0.9675, and the MAPE error decreased from 2.87% to 2.57%. The most significant improvement has been achieved for atmospheric pressure. The R2 metric increased from a negative value of -0.124 to 0.974, and the MAPE error decreased from 0.12% to 0.06%. The results confirm that the proposed pipeline reduces the error and improves the quality of multisensory data, minimizes the amount of data being transferred. The practical value lies in creating a continuous cycle of local filtering and merging of data without accumulation of errors.

The low hardware and software development success is a problem in CAD. The problem entails large expenditures of material, technical and human resources. In order to increase the product development success need to solve the issue of integrating virtual prototypes in the design stage. The issue includes a digital representation that allows evaluate the design's operability. The paper describes the integration between LOTSMAN:PLM and Unity 3D, that includes the modified algorithm development for searching and extracting 3D assemblies in the repository. This algorithm has a name api.getinfo, that allows to aggregate design, technical and operational information. These information puts into a query to extract a 3D assembly for a graphical repository. A specialist can estimate the assembly ergonomics in virtual reality. At industrial enterprise computational experiments were carried out with a modified algorithm implemented like an integrator of 3D assemblies into virtual reality. 3D assembly’s examples were obtained in a graphical repository. A specialist can estimate the location of working bodies and blocks in virtual reality. The integrator interacts with products from ASCON, SolidWorks, AutoCAD, Altium Designer and Delta Design.The modified algorithm reduces the time required to build, evaluate, and edit a 3D assembly by 10% (1.25 business days). Results are based on the experiments. The integrator is embedded in a design and industrial enterprise and in training (advanced training, retraining) of specialists. The work uses a set-theoretic justification for mathematical description of expressions, parameters, etc. Further research is related to the identification of errors in the behavioral specialist model

The relevance of this research is driven by the need to improve the accuracy of human action recognition systems in video by integrating audiovisual information, which compensates for the limitations of each individual modality. The aim of this work is to develop and investigate methods for late fusion of predictions from independently trained video and audio models for recognizing six classes of human actions (Archery, BalanceBeam, BasketballDunk, BoxingSpeedBag, FloorGymnastics, SoccerPenalty).
The study employs 3D convolutional neural networks with Squeeze-and-Excitation blocks for video processing and a hybrid CNN-GRU architecture for audio analysis, along with seven prediction fusion methods: arithmetic mean, weighted average, majority voting, confidence-based fusion, stacking, product of probabilities, and the Borda count. Experiments demonstrate that the arithmetic mean and product of probabilities methods achieve the highest accuracy of 91.67%, outperforming the individual modalities by 2.33% for color and by 5.6% for grayscale video models. Confidence-based fusion exhibited the best results for grayscale models by automatically adapting the contribution of each modality based on the entropy of their predictions. Special attention is paid to the strict separation of training, validation, and test sets based on filming environments to prevent data leakage and to ensure objective evaluation of model generalization capability. The obtained results confirm the practical value of the proposed approach for video surveillance systems and multimedia platforms, offering an optimal balance between accuracy, computational complexity, and robustness to the absence of one modality. This research contributes to the development of multimodal machine learning methods and provides practical guidelines for selecting a fusion strategy depending on available computational resources and input data types.

Intentional jamming represents a serious threat to the security and availability of wireless communication systems. Modern wireless networks, including cognitive radio, sensor networks, and Internet of Things infrastructures, are particularly vulnerable because adversaries can dynamically adapt their jamming strategies. The objective of this study is to develop an adaptive anti-jamming algorithm capable of maintaining communication reliability under dynamic interference conditions. To achieve this goal, the interaction between the legitimate transmitter and the jammer is modeled as a Markov Stackelberg game, where the legitimate node acts as a leader and the jammer acts as a follower. Reinforcement learning is used to determine the optimal strategy of the leader in a stochastic environment, while robustness against channel uncertainty is ensured through a SOCP (Second-order cone programming) formulation that guarantees the required quality-of-service constraints. The learning process is implemented using the SAC (Soft Actor-Critic) algorithm, which enables stable policy optimization in continuous action spaces and stochastic environments.
The research tasks include the formalization of the anti-jamming interaction as a Markov decision process, the integration of reinforcement learning with a robust SOCP optimization layer, and the evaluation of the proposed approach through simulation. A Monte Carlo simulation of the proposed algorithm, as well as several algorithms based on FHSS (Frequency-Hopping Spread Spectrum), was conducted. The proposed anti-jamming algorithm reduces the probability of communication outage by 0.14. The results indicate that the proposed approach improves the resilience of wireless communication systems and reduces the probability of successful denial-of-service attacks at the physical layer

Joint Automated analysis of multiphase CT scans, acquired at different time points after contrast agent administration, is highly important for accurate pathology detection. However, it faces a fundamental problem of spatial misalignment between phases due to patient breathing and movement, which leads to a significant reduction in the accuracy of automatic segmentation. Existing pre-registration (phase alignment) methods require manual parameter tuning and are not integrated into trainable pipelines, hindering the full automation of the segmentation process. The aim of this study was to develop a differentiable method for aligning multiphase CT images based on trainable linear affine transformations, fully embedded into the segmentation model training pipeline. Unlike traditional approaches, the proposed method implements a differentiable affine registration module (including translation, rotation, and scaling), whose parameters are optimized via gradient descent without manual tuning and are directly integrated into the computational graph of the nnU-Net model. The study compared the proposed method with a baseline multiphase segmentation approach (based on simple phase concatenation without registration) and an affine registration method implemented using the ITK.Elastix library on the open abdominal CT dataset WAW-TACE. Compared to segmentation without registration, a substantial improvement in quality metrics was achieved: an increase in the Dice coefficient by 43.19% and in the ROC-AUC metric by 13.55%. Compared to Elastix, the improvements were 13.43% in Dice, 6.86% in ROC-AUC, and 19.6% in the accuracy of pathology count detection per CT scan. The practical significance of the research lies in the development of a ready-to-use PyTorch module for integration into CT image analysis pipelines. It enables fully automated registration without the need for manual hyperparameter tuning and offers high computational efficiency

The relevance of this work is motivated by the fact that in specialized domains and low-resource languages obtaining a sufficient number of “document-summary” pairs for effective abstractive text summarization is expensive and often practically infeasible, whereas modern deep learning models require large labeled corpora and large volumes of text remain unused. The aim of this study is to develop a new training method for abstractive summarization models in low-resource (10-shot and 100-shot) settings that improves quality by fine-tuning the model on unlabeled data using unsupervised contrastive learning with input augmentation. The research tasks include designing a new contrastive loss function and comparing the proposed approach with existing methods of low-resource abstractive summarization. The methods and approaches comprise a new loss function for fine-tuning a transformer that includes a contrastive generative component based on a differentiable approximation of the ROUGE-3 metric. Two variants of the method are proposed – the sequential “DiffROUGE-seq” and the semi-supervised “DiffROUGE-sim”. BART-large is used as the base model, while input augmentations are generated by FLAN-T5-large in a zero-shot regime. Experiments are conducted on popular datasets AESLC, Gigaword, XSum, and Reddit. The proposed method achieves substantial improvements in ROUGE scores in successful cases, with average gains of 2.69 in the 10-shot setting and 2.13 in the 100-shot setting. In summary, the newly proposed low-resource abstractive summarization method that leverages both labeled and unlabeled data is significantly more effective in terms of ROUGE than existing approaches

The paper is devoted to the development of an algorithm for identifying the parameters of a homogeneous nested piecewise linear regression of the first type – a model that is in demand when analyzing complex systems whose behavior cannot be adequately described by smooth functions. The review part of the work systematizes modern publications illustrating the use of piecewise linear forms in various subject areas: from modeling energy consumption and nonlinear control systems to image processing, reconstruction of genetic networks, and filtering of geophysical data. The novelty of the study lies in the fact that for the first time for this class of models an identification algorithm based on the mixed estimation method (MEM) is proposed, which allows flexible combination of two different quality criteria. It is shown that by introducing additional Boolean variables and auxiliary constraints, the original optimization problem is reduced to a standard linear Boolean programming problem, which makes it possible to use available numerical methods for its solution. The effectiveness of the developed algorithm is demonstrated on real data of the mining and metallurgical company Norilsk Nickel for 2010–2024. The dependent variable is revenue, and the predictors are the production volumes of nickel, palladium, copper and platinum. Two alternative models are constructed – using the classical least absolute deviations method and the proposed mixed estimation method. A comparative analysis shows that the second model has a slightly higher average percentage error, but significantly outperforms in the magnitude of the maximum error on the control subsample, which makes it preferable in conditions where outliers are critical. The results obtained confirm the practical value of MEM for constructing interpretable and robust regression dependencies

Digital Human Models (DHMs) are becoming an important tool in robotics, enabling reproducible and controlled validation of computer vision algorithms. Such algorithms include human detection, tracking, and following, which play a key role in autonomous navigation, trajectory planning, and safe robot-human interaction in shared spaces. However, standard simulators are typically limited in the realism of virtual actors, the diversity of their appearance, and the variability of their behavior. These limitations reduce the reliability of the results and make it harder to transfer algorithms to real-world robots. This paper presents a set of customizable DHMs for the Gazebo simulator. The developed models feature variability in clothing, anthropometric parameters, and appearance, and include a library of walking animations simulating various human movement scenarios, including changes in speed and direction. To evaluate the effectiveness of the proposed approach, virtual experiments were conducted in a realistic office environment using the TIAGo Base mobile robot with differential drive, equipped with a 2D lidar and a monocular camera. The efficiency of the user-following algorithm was evaluated in terms of average distance traveled and the number of target loss events (the number of false positives). The results showed that the proposed DHMs make it possible to reproduce complex perception conditions, including occlusions and dynamic obstacles, while providing scalable and systematic validation of algorithms. The obtained data confirm the significance of the developed DHMs for testing robotic systems oriented toward human interaction in conditions close to real-world settings

Relevance. Modern requirements for specialized automated control systems and advances in communication technology are reviving interest in the HF band as a promising means of controlling ground, sea, and air-based robotic complexes (RCs). Existing adaptation algorithms for HF communication systems, by national and international standards are reactive and do not take into account the frequency dependence of ionospheric parameters, which limits their effectiveness for reliable RC control. Objective. To develop an algorithm for selecting the optimal operating frequency and transmission parameters (bandwidth, data rate, power, interleaving depth) based on prediction of frequency-dependent characteristics of a diffuse HF channel using passive GNSS monitoring data. Methods. A structural multipath model for correlation intervals, the ITU-R method for the mean signal-to-interference ratio, and analytical expressions for the Nakagami parameter  and the permissible signal-to-interference ratio taking into account diversity reception are used. Results. An algorithm has been developed that includes estimating ionospheric parameters from GNSS data, calculating the frequency-dependent characteristics of the channel, and selecting the optimal operating mode that maximizes the energy margin and communication reliability. Numerical simulation for a typical path confirmed its efficiency. Conclusions. The proposed algorithm provides proactive selection of communication parameters taking into account the real state of the ionosphere, increasing reliability and data rate, which is critically important for RC control tasks. It is a step towards the creation of cognitive HF systems.

The article is devoted to the consideration of current trends and approaches to solving the urgent optimization problem of three-dimensional bin packing problem. The importance of building effective methods for solving this problem is due to the rapid growth of e-commerce, where achieving even incremental improvements in container filling density can lead to significant reductions in freight transportation and storage costs. The article provides an analysis of various types of problems and suggests a classification of bin packing problems according to various criteria, including: offline and online packing, by dimension, by type and quantity of containers and cargo. The formulation of the classical optimization knapsack problem is given and various options for constraints due to the specifics of the tasks being solved are considered. A brief overview of the main approaches to solving the problem is given. The analysis and generalization of the characteristic features of the application of metaheuristic approaches based on the use of evolutionary and bioinspired algorithms and machine learning methods is carried out, their advantages and disadvantages are noted. Due to the complexity of the problem under consideration, it is proposed to actively use known and develop new modifications of metaheuristic algorithms that make it possible to find quasi-optimal solutions in polynomial time. The analysis of known machine learning methods and bioinspired algorithms is given, the principles of their operation are described, their main features, advantages and disadvantages are highlighted, and the prospects for their development and application to solve NP-complete combinatorial optimization problems are noted. A generalized principle of operation of metaheuristic algorithms is given. A comparative analysis of the application of various optimization methods has shown the effectiveness of using metaheuristic methods to solve the problem of three-dimensional packaging.

An analysis of the literature on the topic of fully homomorphic encryption is carried out. A brief description of the completely homomorphic Domingo-Ferrer cryptographic system based on the number factorization problem is presented, and the time characteristics of the stages of an attack with a known plaintext on this cryptosystem are given. The time characteristics of cryptosystem operations are analyzed, methods and means of their practical implementation are described. New algorithms for implementing the operations of the Domingo-Ferrer cryptosystem are proposed to reduce their execution time. The justification of estimates of the time costs of cryptosystem operations is formed on the basis of theoretical calculations, as well as the results of experimental studies. The aim of the study is to reduce the execution time of the Domingo-Ferrer cryptosystem by developing algorithms for their modification, taking into account the specifics of practical implementation. The main result of this work is a reduction in the execution time of the following operations of the Domingo-Ferrer cryptosystem: encryption by 10-15%, decryption by 2 times, homomorphic multiplication by 64 times for a chain of 200 multiplications using the degree of polynomials of the ciphertext representation d=100 and a slight increase in the time spent on key generation. The conducted research represents a significant contribution to the development of a fully homomorphic Domingo-Ferrer cryptosystem based on the integer factorization problem. This work has practical significance because it significantly improves the performance of homomorphic calculations of this cryptosystem. The results obtained can become the basis for the development of efficient (in terms of required computing costs and the level of security provided) cloud computing software and hardware systems using a fully homomorphic Domingo-Ferrer cryptosystem to ensure the confidentiality of processed information

The automatic extraction of causal relations is critically important for decision support systems, but its development is hindered by the scarcity of annotated corpora. The aim of this study is to comparatively analyze the effectiveness of three types of training data: open expert corpora, a manually annotated political corpus, and synthetic data generated by a large language model. Experiments were conducted using the DistilBERT architecture with token-level BIOES tagging. Synthetic data were generated with the Llama-3 LLM by encapsulating causes and effects in XML markers, followed by deterministic conversion into the BIOES format. For comparison, two expert corpora (EventStoryLine and SemEval-2010 Task 8), the expert-annotated political corpus PolitiCAUSE, and a synthetic sample of 1,100 balanced examples were used. The model trained on synthetic data achieved a Macro F1 score of 0.736, which is 6.7% higher than the result obtained from training on the expert corpora (0.690). Training on the political corpus PolitiCAUSE (24,417 examples) yielded a substantially lower result, with a Macro F1 score of 0.367. This discrepancy is attributed to class imbalance (approximately 15% causal tokens versus 50% in the synthetic sample) and to differences in task formulation: PolitiCAUSE is oriented toward binary sentence-level classification of causal presence, whereas the present study addresses token-level span labeling. The hypothesis that annotation quality and class balance are more important than data volume is confirmed. The proposed LLM-generation pipeline makes it possible to create training datasets that outperform expert corpora without direct manual annotation costs. The method is recommended for rapid deployment of causal relation extraction systems in new domains.

The transition of the cloud, thunderstorm state, accompanied by intensive lightning discharges, to hail occurs as a result of the nucleation and growth of hail nuclei in the drop and grain forms. The vertical extent of the volume where nuclei where are formed on giant particles of micro-impurities in the atmosphere (GPA) is between the isotherms from -2 ⁰C to -15 ⁰C. Ice crystals grow due to sublimation and coagulation with supercooled cloud droplets. In the same volume, between the isotherms of temperatures from -5 ⁰C to -16 ⁰C, discharges occur with the maximum number of lightning strikes at an altitude
of -10 ⁰C. Lightning strikes introduce an excess volume charge with an ion concentration of about 10¹⁷ cm^-3 into the cloud space. Positive lightning strikes introduce positive ions, while negative lightning strikes introduce negative ions. These ions are deposited on the GPA, which have concentrations of about 10⁵ m^-3 in pre-hail situations. The negative charge on the GPA, contributing to the nucleation of planar crystal structures on the particle, forms a granular hailstone. The positive charge on the GPA forms filamentous structures that break off, and the fragments, when they come into contact with supercooled large droplets and freeze them, lead to the formation of droplet hailstones. To prevent hail, ground lightning, and the redistribution of precipitation, a method based on the initiation of artificial lightning has been proposed. Methods of initiating artificial lightning are given using: plasma formations resulting from the detonation of 100 mm artillery shells in thunderstorm-hail clouds, a special rocket with a detonating cord, and an anti-hail rocket with a conductive surface. A significant change in the radar reflectivity of a thunderstorm-hail cloud has been observed under the influence of an artificially initiated lightning strike. Based on the results of preliminary experiments on hail prevention, the method of preventing and interrupting hail using "reagent plus artificially initiated lightning" has been proposed. Given the established fact that approximately 10 tons of precipitation are released with each lightning strike, it is possible to redistribute precipitation, prevent ground-based lightning, and more using artificially initiated lightning. In the future, unmanned aerial systems should be used to initiate artificial lightning within thunderstorm clouds, while quadcopters should be used in the vicinity of clouds

The paper presents a mathematical model of the atmospheric surface electrode layer structure, taking into account the combined action of global and local electric field generators. Spatiotemporal modeling is based on the conjugation of harmonic trigonometric functions of the diurnal cycle for local turbulent exchange and for the global total current density. When modeling the behavior of the electric field, analytical solutions of the electrode effect equations in the atmosphere are used. The proposed approach consists of a step-by-step substitution of the total electric current density and the turbulent diffusion coefficient instantaneous values, obtained from the equations of their periodic variation for a specific hour of the day, into analytical formulas for the stationary spatial distribution of the turbulent electrode layer characteristics. The validity of using the quasi-stationary approximation is strictly substantiated by the significant difference (by more than two orders of magnitude) between the time scales of establishing electrical equilibrium in the medium and the period of the global current generator diurnal variation. It was established that the combined modeling mode adequately reproduces the synergistic expansion of the electric field strength (potential gradient) diurnal curve range and the deformation of its profile, characteristic of the summer season in the high mountains. The reliability of the obtained theoretical results was confirmed by their comparison with experimental data from in-situ measurements at the alpine station of Peak Cheget (430 16' N and 420 30' E), located in the Elbrus region at altitudes of 3040 m above sea level.
The analysis demonstrated good quantitative and chronological agreement between the calculated extremes values and the observed diurnal variations in electric field during the summer season. The results obtained can be directly used to improve the accuracy of geophysical monitoring data interpretation

This paper addresses the problem of resonant excitation of bushes in two-dimensional lattices with discrete symmetry. A bush is defined as a set of vibrational modes that persists over time and constitutes an exact solution to the nonlinear equations of motion, which cannot be obtained within the framework of perturbation theory. The relevance of this study lies in the fact that the proposed scheme for bush excitation, in contrast to previous mathematical models aimed at exciting bushes of nonlinear normal modes, can be implemented in a physical experiment. The above-mentioned ensembles of vibrational modes are considered within the framework of the theory of Chechin and Sakhnenko, based on group-theoretic research methods, which makes it possible to use any models of interatomic interaction potentials with equal success. This, in turn, allows us to consider dynamic systems of multiple scales, from molecular vibrations to the movement of macroscopic crystal structures. The aim of the work is to present a mathematical model of excitation of stable nonlinear dynamic objects in two-dimensional molecular structures suitable for both computational and physical experiments. Providing the results of a computational experiment with a description of the detected dynamic objects. Unlike traditional methods, Bush theory provides a solid theoretical framework based on group theory that overcomes the limitations of perturbation theory, offering results of exceptional reliability. At the same time, the symmetry properties used in Bush theory can provide significant advantages for numerical modeling, significantly reducing computational complexity. Now, this issue is being investigated by many authors around the world. It is worth noting, however, that the works in which the problems of excitation of dynamic objects close to the concept of bouches are discussed relate only to one-dimensional cases and that group-theoretic methods are not considered in them, with very rare exceptions

Existing electronic voting protocols often feature an architecture where ballot submission is an irreversible action. This significantly reduces system usability, depriving the participant of the right to correct an accidental error or change their opinion during lengthy decision-making procedures. The aim of the work is to develop a model of a cryptographic protocol that provides a native capability for multiple decision changes to enhance usability while maintaining strict anonymity and counting correctness.
The architecture is based on the integration of a trapdoor commitment scheme to implement choice flexibility, a verifiable random function (VRF) for ballot generation, the Privacy Pass protocol for anonymous authorization, and mix-networks. An electronic voting protocol has been developed that implements the possibility for a participant to change their decision multiple times. The fulfillment of basic and some extended security properties, including, but not limited to, participant confidentiality and universal verifiability, has been theoretically proven. It is shown that, unlike existing analogs, changing a decision in the proposed model does not require repeating a complex authentication procedure or key revocation, which minimizes the number of interactive user actions. The proposed solution aligns the user experience of secure voting with the standards of modern digital services by providing a choice editing function, thereby increasing the convenience of the voting procedure for participants. The modular structure of the model allows for the application of any robust cryptographic algorithms, including post-quantum ones

The influence of delays in receiving navigation information on maintaining a formation of underwater vehicles in the leader-followers mode is studied. The study is based on the assumption of a gradient principle of formation control during movement. The complexity of the problem is due to the unknown trajectory of the leader, incomplete knowledge and non-stationarity of the environment, and measurement errors. The obtained results are independent of the configuration of vehicles in the formation. It is shown that it is sufficient to study the influence of navigation delays on the motion of a single vehicle. Key assumptions: the trajectory of the leader is unknown; all interactions between the leader and follower vehicles are synchronized; the leader transmits its coordinates to the other vehicles at a certain frequency; displacements of the follower from its position in the formation can occur in any spatial direction. It is assumed that between receiving data on the leader's position, the follower vehicle moves in a straight line in a direction that can be chosen in many ways. One option for choosing such a direction is considered, and a mathematical model of the motion is constructed. Since the moment of information receipt is unknown, a nonlinear law is used to describe the follower's motion, taking into account the distance to the target point. Estimates of the vehicle's deviation from the target position in the formation at each moment in time are found, as well as average estimates of its deviation over the time interval between receiving navigation information and over the entire trajectory. An upper bound for the integral deviation is proposed and substantiated, allowing for a simple interpretation of the factors influencing deviation and formation breakdown.

The key unit of kraft pulp production – the recovery boiler – ensures the regeneration of black liquor chemicals. The task of increasing the liquor reduction degree without changing the process technology or equipment design, i.e., through information and algorithmic means – automatic control is relevant. A higher guaranteed quality of recovery is provided by control algorithms synthesized on the basis of adequate mathematical models. The aim of this paper is a system analysis of the recovery boiler as a control plant and the development of an enhanced mathematical model suitable for synthesizing high-quality automatic stabilization systems. The methodology is based on the principle of sequential uncertainty removal for complex control system models (A.A. Vavilov). Unlike known approaches, the dynamics of the recovery boiler in the vicinity of a selected operating point are described as a multi-level LTI-class model. Combining analytical methods with experimental data processing for a specific operating mode made it possible to obtain mathematical models of an actual recovery boiler with parameters determined down to specific numerical values. The developed multilevel linear stationary model of the recovery boiler makes it possible to analyze the dynamics of subsystems and the unit as a whole, the stability of modes and the influence of disturbances, and justify the need to automatically maintain the required quality of technological processes in the vicinity of the selected mode. Symbolic models are applicable to most similar kraft process units. Parameters of fully defined models should be adjusted based on the results of processing data related to a specific unit and mode. The direction of further research is associated with the development of methods for analyzing the dynamics of complex control systems according to models of a hierarchical structure, as well as the synthesis of systems for stabilizing the regime and reducing the influence of perturbations.ls, as well as the synthesis of systems for regime stabilization and mitigation of the influence of disturbances

The paper presents the concept of adaptive control in natural gas separation. The control systems used in practice are usually based on fixed control algorithms and do not consider dynamic changes in physical, technical and operational parameters, which leads to a decrease in separation quality and an increase in energy consumption. The article describes the approach of combining system analysis, modeling and the Pareto method. This paper presents a concept for adaptive control in natural gas separation. Control systems used in practice are typically based on fixed control algorithms and do not account for dynamic changes in physical, technical, and operational parameters, which leads to reduced separation quality and increased energy costs. This paper describes an approach combining systems analysis and the Pareto method. The aim of this paper is to develop a concept for adaptive control of the natural gas treatment process based on systems analysis. Recent studies demonstrate that data-driven methods enable more precise parameter adjustment, better responsiveness to raw-gas fluctuations and improved impurity removal efficiency. To achieve this goal, this paper addresses the challenges of identifying and classifying factors affecting the quality and efficiency of gas separation, as well as integrating their relationships within a unified conceptual control model. Particular attention is paid to the impact of precise control of key parameters, such as pressure, temperature, and flow rate, on the efficiency of these processes. An analysis of recent research demonstrates the growing use of neural networks and machine learning models in gas purification for predictive control, anomaly detection, and optimization of operating parameters. A comparative evaluation of classic PID controllers, fuzzy, adaptive, and neural control methods confirms the advantages of intelligent control in terms of stability, adaptability, and energy efficiency. The main result of this study is the substantiation of key factors determining separation efficiency, among which pressure, temperature, and gas flow rate have the greatest impact. The resulting model forms a methodological basis for the development of intelligent and adaptive control systems for gas purification processes.

The article presents the developed matrix method of simultaneous assessment of the reliability of several variants of complex multi-agent reconfigurable systems, as well as the staged method of reservation, which differ from the known ones in their simplicity and wide scope of application The matrix reliability assessment method consists in a matrix display of the reliability parameters of all agents of a specific variant of a group of system agents and a vector representation of the mission conditions. It is assumed that the mission reliability of a target group of agents is determined solely by the reliability of the agents themselves and the nature of their redundancy. The staged redundancy method is distinguished by its use of traditional standard redundancy schemes as stages in the process of redundant agent groups, significantly increasing the reliability of system variants. System agents are divided into active agents, directly performing the current mission, and passive agents, awaiting activation. Active agents, in turn, are divided into primary and backup agents. Primary agents initiate mission execution, while backup agents, awaiting replacement of failed active agents, can be in either an active state (hot standby) or a cold standby state. Specific mission conditions are taken into account by a vector of correction factors, significantly improving the accuracy of reliability assessment and agent group uptime prediction. An important feature of the proposed methods for assessing and predicting reliability is their analytical nature, which allows the use of modern high-speed computing tools in solving problems of ensuring the required reliability of missions performed by complex multi-agent systems

A 34-year archive (1990–2023) of radiosonde data from 11 stations in southern Russia, ERA5 reanalysis is used to analyse mid-tropospheric dry intrusions (700–500 hPa) as a predictor of hazardous convective downdrafts, downbursts and microbursts. Separate samples of 268 verified cases of severe convective squalls (wind ≥ 25 m s–1) and 1042 control thunderstorm days without damaging wind are compiled. Mean dewpoint depression at 700 hPa is 2.4 times higher in squall days than in controls
(18.3 vs. 7.7°C), whereas CAPE shows no statistically significant difference between the two groups.
A novel diagnostic parameter – Dry-Enhanced Downdraft CAPE (DCAPEdry) – is derived as the product of standard DCAPE and the normalised dewpoint deficit in the 700–500 hPa layer. DCAPEdry ≥ 620 J kg–1 achieves POD = 84% and FAR = 19% in squall detection, markedly outperforming DCAPE alone
(POD = 61%, FAR = 43%). The roles of evaporative cooling, surface vortex generation, and Ciscaucasian orography in downdraft intensification are discussed. Orographic features (valley focusing effect, foehn zones, and convergent zones at the border between mountains and plains) systematically increase the intensity of descending flows, contributing to the concentration of the most dangerous squalls in the area from Nalchik to Stavropol.

Uncertain and imprecise data are typical for multicriteria problems, making fuzzy set theory an adequate tool for their solution. The objective of this paper is to apply the TOPSIS method in a fuzzy environment using syntactically independent linguistic variables. Due to the frequent occurrence of fuzzy concepts in decision-making data, crisp values are insufficient for modelling real-world situations. In the proposed approach, the evaluation of each alternative and the weight of each criterion are expressed by syntactically independent linguistic variables, whose arbitrary values are determined through the semantics of the base terms. An approach to calculating arbitrary (non-base) values of linguistic variables used by an expert is considered. The base values of the syntactically independent linguistic variables used are specified by triangular fuzzy numbers. The proposed approach is well suited for solving group decision-making problems in a fuzzy environment, where an expert or group of experts is not constrained by a limited number of fuzzy assessments. Here, fuzzy variables representing the values of the syntactically independent linguistic variables "criterion importance" and "criterion score" are considered as weighting coefficients for criterion importance and evaluation of qualitative criteria. The paper proposes using the centre of gravity method to calculate the distance between two triangular fuzzy numbers. In accordance with the fuzzy TOPSIS concept, a proximity coefficient is defined to determine the ranking order of all alternatives by calculating the distances to the fuzzy positive ideal solution (PIS) and the fuzzy negative ideal solution (NIS). An example is provided to illustrate the proposed approach

Relevance. Effective mathematical models for organizing parallel computing utilize principles of simultaneous rule execution on independent data fragments, as well as methods for processing symbolic information as a unifying category for various data types. Production systems in A.A. Markov Jr.'s notation offer the necessary potential for creating such models, but their standard control scheme is sequential, limiting their application in applied problems. Therefore, synthesizing modified production systems and their control schemes aimed at parallelizing the computational process of processing symbolic information is a pressing issue. The goal of this study is to reduce the execution time of a modified production model through bidirectional processing of symbol strings. The solution method is based on the application of principles of bidirectional grammatical parsing from the theory of syntactic analysis and compilation in algorithmic production systems and consists of modifying A.A. Markov's standard production computation scheme. This is achieved by introducing single-shot productions, bidirectional access to the string being processed, decomposing the original system into sections, and introducing a set of guard conditions for conflict resolution. The production model is implemented for typical symbolic processing tasks: word tagging and word reversal. Results. The structure of a modified production system was developed that overcomes the limitation of A.A. Markov's standard scheme—the sequential nature of computations. As a result of the study, a formal model of serial-parallel processing was created, a new control scheme was developed, and the correctness of the computations was demonstrated. Simulation on typical tasks showed a reduction in word processing time by almost half, achieved, in part, by eliminating the mandatory return to the first production in the control scheme and aggregating productions into sections. Conclusion. The study confirmed the feasibility and effectiveness of transforming A.A. Markov's sequential production systems into serial-parallel production models that produce correct results. A promising area of application is methods and software and hardware in homogeneous computing systems for high-performance processing of symbolic information