РУС
ENG
№ 1(40)'2026
Aims and Scope
Founder and Publisher
Editorial board
Licensing terms
Privacy Statement
Plagiarism policy
Publication ethics
Archiving Policy
Subscription
Instructions for authors
Reviewing proccess
Copyright Notice
License agreement
Article processing charges
All issues
Search
Contacts
Section
ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING
Emelianov V. A., Chernyi S. G.,
Emelianova N. Yu., Kochkarov A. A., Prokopchina S. V.
RECOMMENDATION SYSTEM FOR DETERMINING THE METAL PROPERTIES BASED ON NEURAL NETWORK
UDC 004.032.026
DOI 10.24412/2413-7383-2026-1-40-4-14
Language: English
Annotation: The paper describes a recommendation system that is designed to determine the properties of metals.
A structure of an expert subsystem is developed which designed to determine the nominal values of the
properties of metals. A fragment of the knowledge base of metals, which contains a set of rules on their
properties, is shown. Neural network is developed a to refine the values of the properties of metals. The
results of simulations of the developed neural network are shown.
Keywords: neural networks, metal properties, expert subsystem, classification.
Kamguia Feukwi Herve, Goncharova А. B.
SYMMETRY LINE DETECTION IN IMAGES USING ANGLED-LINE AND MIDDLE-OUTWARD TECHNIQUES
UDC 004.932
DOI 10.24412/2413-7383-2026-1-40-15-24
Language: English
Annotation: Reflection symmetry is a key geometric feature that plays a vital role in various fields like computer vision,
pattern recognition, and medical image analysis. In this research, we present a refined and easy-tounderstand approach for detecting symmetry lines. We use two different techniques: a straightforward bruteforce angled-line sweep and a clever middle-outward search strategy. These methods leverage reflection
transformations, assess the similarity of binary masks, and rely on a solid mathematical foundation.
Keywords: symmetry; reflection; rotation; images; Jaccard coefficient; cosine coefficient;
explainable artificial intelligence (XAI).
Sichkar Valentin N.
ADVANCED CONTEXTUAL RECONSTRUCTION ARCHITECTURE FOR ROBUST OBJECT DETECTION UNDER PARTIAL OCCLUSION
UDC 004.855.5
DOI 10.24412/2413-7383-2026-1-40-25-40
Language: English
Annotation: Partial occlusion causes critical performance degradation in object detection: accuracy drops from
92-94% to 65-71% mAP@0.5 when occlusion exceeds 40% of object area, despite 60-80% of realworld objects experiencing occlusion. We present ACRAFD (Architecture for Contextual
Reconstruction and Adaptive Fragment-aware Detection), integrating contextual reconstruction,
fragment-aware analysis, and adaptive attention. The architecture incorporates Unified ContextFragment Module (UCFM), Hybrid Adaptive Attention Block (HAAB) with IoU-aware weighting, and
Feature Completion Layer (FCL) based on diffusion models. We introduce Fragment-Aware Focal
Loss and specialized metrics: FA-mAP, COS, RS. Experiments on MS COCO 2017 demonstrate:
ACRAFD achieves 93.2% mAP@0.5 on heavily occluded objects (+21.4 percentage points over
baselines), FA-mAP 81.7% (+22.6%), COS 0.91 (+17.0%), RS 94.8% (+11.3%). Results show
substantial improvements for safety-critical computer vision applications
Keywords: ACRAFD, object detection, partial occlusion, deep learning
Dzeboev D. I.
DIVISION IN TENSORIAL ALGEBRA
UDC 512.554.1
DOI 110.24412/2413-7383-2026-1-40-41-49
Language: Russian
Annotation: The paper investigates a consistent definition of division in tensorial algebra, a generalized structure
where multiplication is defined by a third-order tensor and may be noncommutative and
nonassociative. It is shown that the classical definition via an inverse element applies only to
associative unital algebras, whereas in the nonassociative case division must be formulated as the
solvability of left and right multiplication equations and is therefore partial and potentially non-unique.
In tensorial algebra, this problem naturally reduces to solving systems of linear equations arising from
interpreting a tensorial element as a linear operator, leading to a unified conceptual framework of
division compatible with both associative and nonassociative structures.
Keywords: tensorial algebra, division, inverse element, associativity, nonassociative
algebras, left and right division, linear operator.
V. I. Lyikin, D. V. Lobov, D. V. Loginov
TREE TRUNK RECOGNITION USING FOUR MACHINE LEARNING MODELS
UDC 004.932.2
DOI 10.24412/2413-7383-2026-1-40-50-62
Language: Russian
Annotation: This article investigates the applicability of modern neural network architectures (Faster R-CNN,
YOLOv11x, DETR, RetinaNet) for the automatic detection of tree trunks in the forest conditions of the
Republic of Karelia. A specialized dataset was compiled and annotated for the study, combining open
data and original images adapted to the specifics of northern forest ecosystems. Experiments established
that the Faster R-CNN model demonstrates the highest detection accuracy (mAP@0.5 = 0.659), while
YOLOv11x provides the optimal balance between accuracy and processing speed (15.33 FPS), which is
critical for real-time systems. The results confirm the practical feasibility of automating forest inventory and
monitoring tasks using computer vision methods and indicate the need for further dataset expansion to
improve the models' generalization capability.
Keywords: artificial intelligence, machine vision, object recognition, machine learning
models, forestry complex.
I. I. Maksimenko
IDENTIFICATION OF SPECIAL ALGEBRAS IN TERMS OF BAERE AND P-ADIC METRICS
UDC 519.713.4
DOI 10.24412/2413-7383-2026-1-40-63-74
Language: Russian
Annotation: The article studies the process of identification of objects of a potentially infinite class with a standard
for various automata (synchronous, asynchronous and generalized automata) and non-automata
structures (unstructured sets and lattices) based on a generalized representation in terms of a special
type of "Baire" metric. It is shown that similar results are valid for p-adic metrics. Criteria for the
existence of representations in terms of the properties of limit objects of the class are obtained. This
criterion draws parallels between the process of identification of an object with a standard and the
properties of limit points of special metric spaces.
Keywords: automaton, lattice, identification, representation, metric, fragment,
cofgranment, limit objects.
K. I. Morev
UNIVERSAL BACKBONE FOR SOLVING COMPUTER VISION TASKS
UDC 004.932.2
DOI 10.24412/2413-7383-2026-1-40-75-88
Language: Russian
Annotation: The paper explores the possibility of using a single backbone to solve several computer vision tasks
– detecting and classifying objects, extracting descriptors of image keypoints. The YOLOv5s,
MobileNetV2, and ResNet-50 architectures are being investigated as models providing a backbone
for solving all the mentioned tasks. Experiments show that the use of a common feature generator
can significantly speed up the learning and operation of computer vision algorithms, while
maintaining the quality of tasks solving.
Keywords: computer vision, universal backbone, neural network architectures,
image keypoints, keypoints matching.
Topper A. M., Goncharova А. B.
IMPROVING THE STRUCTURING OF CONCEPTS IN LANGUAGE MODELS THROUGH DOMAIN ADAPTATION
UDC 004.932.2
DOI 10.24412/2413-7383-2026-1-40-89-100
Language: Russian
Annotation: The relevance of the study is determined by the need to interpret semantic representations in
language models, especially in the medical field. The Semgeom library was developed for analyzing
embedding geometry using the semantic axes method. A comparative analysis of RuBioRoBERTa
and ruRoberta-large models showed the advantage of the domain-adapted model in the accuracy of
semantic structuring of medical concepts. The results demonstrate the practical value of the method
for validating language models in professional domains.
Keywords: language models; semantic axes; domain adaptation; medical embeddings; AI
interpretability
Section
THEORETICAL INFORMATICS, CYBERNETICS
Yu. Zhang
CLASSIFICATION OF TEXT EMOTIONS BASED ON THE BERT MODEL
UDC 004.932.2
DOI 10.24412/2413-7383-2026-1-40-102-111
Language: Russian
Annotation: The relevance of the study is determined by the need to interpret semantic representations in
language models, especially in the medical field. The Semgeom library was developed for analyzing
embedding geometry using the semantic axes method. A comparative analysis of RuBioRoBERTa
and ruRoberta-large models showed the advantage of the domain-adapted model in the accuracy of
semantic structuring of medical concepts. The results demonstrate the practical value of the method
for validating language models in professional domains.
Keywords: language models; semantic axes; domain adaptation; medical embeddings; AI
interpretability
Section
SYSTEM ANALYSIS, MANAGEMENT AND INFORMATION PROCESSING, STATISTICS
Hamzaoui O. El, Patrakov E. V.,
Borisov V. I.
INTEGRATION OF ARTIFICIAL INTELLIGENCE
AND MACHINE LEARNING METHODS FOR AUTISM
SPECTRUM DISORDER DETECTION: STATE-OF-THE-ART
WITH MULTIMODAL DATA PROCESSING SOLUTIONS
UDC 004.891.3
DOI 10.24412/2413-7383-2026-1-40-113-123
Language: English
Annotation: This review examines the integration of artificial intelligence (AI) and machine learning approaches
for autism spectrum disorder (ASD) detection through multimodal data analysis. ASD, affecting 1%
of the population globally, presents complex diagnostic challenges requiring sophisticated
assessment tools. We analyze state-of-the-art deep learning architectures including CNNs,
RNNs, transformers, and hybrid models applied to neuroimaging, facial expression recognition, and
behavioral analysis. Our analysis reveals that multimodal approaches combining fMRI, video
analysis, and behavioral data achieve superior diagnostic accuracy (85-95%) compared to singlemodality methods. This work provides a comprehensive foundation for advancing AI-driven ASD
assessment methodologies.
Keywords: autism spectrum disorder; artificial intelligence; machine learning; parent-child
interactions; multimodal data integration; deep learning.
Averin G.V., Zviagintseva., A.V., Brazhnikov A.R.
VECTOR AND STATISTICAL CRITERIA
OF RELATIVE COMPLEXITY WHEN COMPARING
OBJECTS BY A SET OF PARAMETERS
UDC 510.52:303.732.4:519.25
DOI 10.24412/2413-7383-2026-1-40-124-136
Language: Russian
Annotation: This article substantiates an approach to assessing relative complexity based on representing
system models in multidimensional state spaces. The empirical basis for this research is datasets
and datasets in biology and artificial intelligence. As an example, the complexity of vertebrates is
analyzed in comparison with humans, the most highly developed organism. Criteria are proposed for
assessing the relative complexity of object states relative to the state of an object adopted as a
standard. Complexity for different classes (groups) of objects is estimated based on the statistical
distributions of the analyzed and standard groups.
Keywords: relative complexity, multidimensional systems, probabilistic and statistical
complexity criteria, integrated measurements and assessment.
Butov O. A., Solod V. S., Nikitina A. A.
INTELLIGENT CONTROL SYSTEM FOR THERMAL HARDENING OF REINFORCEMENT BASED ON A DIGITAL TWIN
UDC 681.518.25/621.771.09
DOI 10.24412/2413-7383-2026-1-40-137-149
Language: Russian
Annotation: The aim of this study is to evaluate the feasibility of creating a digital twin (DT) for managing and
monitoring the thermal hardening process of reinforcing bars based on new, original physical and
mathematical models of accelerated cooling. Based on a review of publications, the objectives of the
DT development, composition, and features are determined. The stages of creating a DT for the
accelerated cooling process of reinforcing sections are described. The DT structure and a program
for monitoring the actual rolling-cooling process are developed using a model for optimizing thermal
hardening modes and emulating rolling mill operation.
Keywords: accelerated cooling, reinforcement profile, digital twin, algorithm, optimization
program, yield strength
Eliseeva N. V., Petrov V. E.
ISTYLISTIC ADAPTATION TECHNOLOGY OF NEURAL MACHINE TRANSLATION
UDC 004.85
DOI 10.24412/2413-7383-2026-1-40-150-158
Language: Russian
Annotation: Neural machine translation is a key technology for automated text translation. Large corporations are
actively integrating neural machine translation technologies into their corporate processes, using them for
global communications and content localization. As these technologies are implemented and used, users
face new demands on machine translation quality that go beyond the accurate rendering of content. This
article explores methods and technologies for improving the quality of neural machine translation. The
technology proposed in the article and the results of testing demonstrate that its use more accurately
conveys the meaning and vocabulary of original phrases.
Keywords: neural machine translation, attention mechanism, tonality of the text, singular
value decomposition, optimization.
Kolesnikov A. E., Martynenko T. V., Shuvatova E. A.
COMPARATIVE ANALYSIS OF SURROGATE-ASSISTED OPTIMIZATION METHODS FOR MICROSERVICE
ARCHITECTURES BASED ON AN INTEGRAL EFFICIENCY CRITERION
UDC 004.852
DOI 10.24412/2413-7383-2026-1-40-159-167
Language: Russian
Annotation: The paper presents a comparative analysis of five surrogate-assisted optimization methods for
microservice architecture configurations: HGP-MOO, RL-MOO, sLQ-EVO, sPERT-EVO and sNSGA-II.
The methods were evaluated by the accuracy of surrogate models for predicting key characteristics and
the quality of the constructed Pareto fronts. The study is based on the integral efficiency criterion E(X)
and a feature space that includes a binary matrix of business function distribution across services and
eight key characteristics of performance, reliability, resource efficiency and structural complexity. The aim
of the study is to justify the choice of the most suitable method for integration into an automated system
for assessment and optimization of microservice architectures.
Keywords: microservice architecture, optimization, surrogate models, integral efficiency
criterion, Pareto front.
Naumov M. A., Voluyeva O.S.
A METHOD FOR DETERMINING GEOMETRIC
PARAMETERS OF MINING FOR AUTOMATED CONTROL
OF MINING EQUIPMENT
UDC 681.586
DOI 10.24412/2413-7383-2026-1-40-168-179
Language: Russian
Annotation: The article is devoted to the development of a methodology for the complex determination of the
geometric parameters of a mine work (drift) and the spatial position of a mining machine (MM). An
integrative computational approach is proposed based on the synthesis of data from an inertial
navigation system and a laser rangefinder system installed on the machine. Algorithms for calculating
key operational parameters are presented, including the actual width of the drift, the azimuth and
slope of its centerline, and the lateral displacement and rotation angle of the MM. The technique is
aimed at providing high-precision navigation and effective control of the combine in real time to
minimize risks and meet project standards.
Keywords: mining combine, geometric parameters of the drift, navigation in mine workings,
equipment positioning, data processing algorithm, integrative approach, real-time control system.
Ostapushchenko D. L., Chernykh V. V., Petrenko M. O., Chalaya E. Y.
DEVELOPMENT OF INTELLIGENT CONTROL SYSTEMS
FOR UNMANNED AERIAL VEHICLES USING
THE PONTREGIN MAXIMUM NEURAL NETWORK
PRINCIPLE
UDC 533.65.622+004.8
DOI 10.24412/2413-7383-2026-1-40-180-189
Language: Russian
Annotation: The work is devoted to the creation of intelligent control systems for unmanned aerial vehicles using
the Pontryagin maximum neural network principle. The problem of constructing a neural network
approximator of a synthetic control function that ensures the movement of an unmanned aerial
vehicle along an optimal trajectory according to given criteria is solved.
Keywords: unmanned aerial vehicle, intelligent control systems, Pontryagin maximum
principle, artificial neural networks.
Pomblas R.A., Shkodyak K.E., Berberova M.A., Varlamov O.O., Adamova L.E.
MIVAR DECISION SUPPORT SYSTEM FOR GAS PLANT STAFF
UDC 004.891+007.52+681.518+65.011.56
DOI 10.24412/2413-7383-2026-1-40-190-201
Language: Russian
Annotation: A mivar expert system (MES) has been developed to support decision-making for gas plant personnel.
The gas plant's compressor units are designed to transport and distribute natural gas for use in the
industrial, commercial, and energy sectors, as well as in oil and gas production operations. The scientific
innovation lies in the creation of a mathematical model of 81 mivar network rules for information
processing and decision-making processes for 10 duty shift personnel at the gas plant. The MES can be
used both to support personnel decision-making and for training or testing.
Keywords: mivar, artificial intelligence, mivar expert systems, smart manufacturing
systems, gas plant compressor units, Wi!Mi, Big Knowledge, automated learning,
mechanical engineering AI.
Reshetnikov K. I., Glukhov M. S., Ronkin M. V.
INVESTIGATION OF COMPUTER VISION OBJECT DETECTION APPROACHES IN ROCK FRAGMENTATION PROBLEMS
UDC 004.932
DOI 10.24412/2413-7383-2026-1-40-202-214
Language: Russian
Annotation :The rock fragmentation problem in open pits is explored using computer vision techniques. The benefits
of employing modern, YOLO-based object detection architectures for both object detection and oriented
object detection are demonstrated, in comparison to traditionally used methods based on Mask-R-CNN
instance segmentation. It is shown that, with the appropriate selection of a confidence threshold for
detection, the performance by the mean average precision (mAP) metric, is statistically indistinguishable.
The threshold may be selected by minimizing the Wasserstein distance between the annotated and the
obtained values using the proposed algorithm.
Keywords: rock fragmentation, computer vision, deep neural networks, object detection, blast
quality estimation.
Trushkin M. I., Krivodubsky O. A.
ANALYSIS OF OPPORTUNITIES TO IMPROVE THE QUALITY OF CUSTOMER SERVICE IN THE TELEPHONE NETWORK
UDC 519.8: 621.391
DOI 10.24412/2413-7383-2026-1-40-215-225
Language: Russian
Annotation:The article examines the concept of quality of service in mobile telephone networks, reviews
existing methods for improving service quality, analyzes limitations of current approaches to quality
management, and outlines requirements for future research.
Keywords: mobile telephone network; base stations; mobile communications; quality of
service; automated optimization.
Наполнение продолжается...
