Journal influence
Higher Attestation Commission (VAK) - К1 quartile
Russian Science Citation Index (RSCI)
Bookmark
Next issue
№2
Publication date:
14 June 2026
Journal articles №1 2026
11. Accelerating the software testing process using continuous integration
[№1 за 2026 год]Authors: Kochkin, D.V. (kochkindv@bk.ru) - Vologda State University (Associate Professor), Ph.D; Krasnov, A.A. (krasnovanton@vk.com) - MMTR Technologies LLC (Engineer); Bogdanov, D.A. (bogdanovda@vogu35.ru) - Vologda State University; Frolov, S.A. (frolovsa@vogu35.ru) - Vologda State University (Senior Lecturer);
Abstract: Using a web application for a procurement information system as an example, this article examines the software testing process within continuous integration (CI). This stage is executed repeatedly and demands significant computational resources. Consequently, reducing testing time to accelerate the development process is a pertinent challenge. The article investigates various approaches to addressing this challenge, including test preparation and test plan formulation. The method of selective regression test execution is considered. This paper proposes applying a multi-agent approach: software agents are responsible for launching tests, analyzing their results, generating recommendations for adjusting the development process, and accelerating the testing process. Time reduction can be achieved by decreasing the execution frequency of individual tests, which should have the lowest probability of containing errors. This probability is estimated based on testing results obtained in previous stages of the continuous integration process. A formal problem statement is provided using a set-theoretic description of the multi-agent system. In the formal model, tests are grouped according to the different requirements for the software product under development. Simulation modeling of the testing process within CI was performed. The results are compared with those obtained using other methods of selective regression test execution. The conclusion is drawn regarding the applicability of the developed approach for accelerating the testing process and reducing the time expenditure on test execution. The novelty and originality of the work lie in the application of the multi-agent approach and in the method for selective test execution based on testing history, which is independent of the programming language and technologies used.
Keywords: testing, the software, software, multiagents systems, control management, continuous integration, selective test execution
Visitors: 3352
11. Accelerating the software testing process using continuous integration
[№1 за 2026 год]
Visitors: 3352
Using a web application for a procurement information system as an example, this article examines the software testing process within continuous integration (CI). This stage is executed repeatedly and demands significant computational resources. Consequently, reducing testing time to accelerate the development process is a pertinent challenge. The article investigates various approaches to addressing this challenge, including test preparation and test plan formulation. The method of selective regression test execution is considered. This paper proposes applying a multi-agent approach: software agents are responsible for launching tests, analyzing their results, generating recommendations for adjusting the development process, and accelerating the testing process. Time reduction can be achieved by decreasing the execution frequency of individual tests, which should have the lowest probability of containing errors. This probability is estimated based on testing results obtained in previous stages of the continuous integration process. A formal problem statement is provided using a set-theoretic description of the multi-agent system. In the formal model, tests are grouped according to the different requirements for the software product under development. Simulation modeling of the testing process within CI was performed. The results are compared with those obtained using other methods of selective regression test execution. The conclusion is drawn regarding the applicability of the developed approach for accelerating the testing process and reducing the time expenditure on test execution. The novelty and originality of the work lie in the application of the multi-agent approach and in the method for selective test execution based on testing history, which is independent of the programming language and technologies used.
Kochkin, D.V. (kochkindv@bk.ru) - Vologda State University (Associate Professor), Ph.D; Krasnov, A.A. (krasnovanton@vk.com) - MMTR Technologies LLC (Engineer); Bogdanov, D.A. (bogdanovda@vogu35.ru) - Vologda State University; Frolov, S.A. (frolovsa@vogu35.ru) - Vologda State University (Senior Lecturer);
12. Emotion recognition using ensemble methods for multimodal sentiment analysis
[№1 за 2026 год]Authors: Fazulyanov, D.V. (fazulianov.dmitrii@gmail.com) - National Research Nuclear University “MEPhI” (Moscow Engineering Physics Institute) (Teaching Assistant); Guseva, A.I. (aiguseva@mephi.ru) - National Research Nuclear University “MEPhI” (Moscow Engineering Physics Institute) (Professor), Ph.D;
Abstract: This article presents an approach for recognizing human emotional states based on multimodal sentiment analysis, employing ensemble methods to integrate data from various modalities (text, audio, and video). Despite progress in this field, existing approaches (e.g., TFN, MARN) have significant limitations related to low robustness against noisy data, substantial computational costs, and difficulties in adapting to the individual characteristics of different modalities. The novelty of this ensemble method lies in the combination of three levels of integration: combining different data types (text, audio, video), diverse models, and ensemble strategies (Stacking, Bagging, AdaBoost, CatBoost), which substantially distinguishes it from known multimodal solutions. The proposed method allows for the consideration of the individual contribution of each modality and minimizes errors from individual classifiers, thereby enhancing the stability and accuracy of the final predictions. The practical significance of the work lies in the broad applicability of the developed approach for automated analysis of message sentiment and user reactions, including social media monitoring, market research, customer satisfaction analysis, and medical diagnostics. A computational experiment was conducted using the open multimodal dataset eNTERFACE'05, containing video recordings of participants' emotional reactions. The best results were obtained using the Stacking strategy (Accuracy = 89.7 %, F1-Score = 89.8 %), which outperforms traditional multimodal models such as TFN and MARN by 10 % and 15 %, respectively, in accuracy, and by 15 % and 17 % in F1-Score. Special attention was paid to studying the resilience of the proposed methods to noise and distortions in input data. The obtained results demonstrated that the developed method maintains high effectiveness and accuracy even under significant noise levels (up to 30 %), substantially expanding its practical application potential.
Keywords: sentiment analysis, multimodal data, ensemble models, Stacking, Bagging, AdaBoost, CatBoost, deep learning, transform-ers, emotional valence, multimodal data, ensemble models, stacking, Bagging, AdaBoost, CatBoost, deep learning, transformers, emotional valence
Visitors: 2766
12. Emotion recognition using ensemble methods for multimodal sentiment analysis
[№1 за 2026 год]
Visitors: 2766
This article presents an approach for recognizing human emotional states based on multimodal sentiment analysis, employing ensemble methods to integrate data from various modalities (text, audio, and video). Despite progress in this field, existing approaches (e.g., TFN, MARN) have significant limitations related to low robustness against noisy data, substantial computational costs, and difficulties in adapting to the individual characteristics of different modalities. The novelty of this ensemble method lies in the combination of three levels of integration: combining different data types (text, audio, video), diverse models, and ensemble strategies (Stacking, Bagging, AdaBoost, CatBoost), which substantially distinguishes it from known multimodal solutions. The proposed method allows for the consideration of the individual contribution of each modality and minimizes errors from individual classifiers, thereby enhancing the stability and accuracy of the final predictions. The practical significance of the work lies in the broad applicability of the developed approach for automated analysis of message sentiment and user reactions, including social media monitoring, market research, customer satisfaction analysis, and medical diagnostics. A computational experiment was conducted using the open multimodal dataset eNTERFACE'05, containing video recordings of participants' emotional reactions. The best results were obtained using the Stacking strategy (Accuracy = 89.7 %, F1-Score = 89.8 %), which outperforms traditional multimodal models such as TFN and MARN by 10 % and 15 %, respectively, in accuracy, and by 15 % and 17 % in F1-Score. Special attention was paid to studying the resilience of the proposed methods to noise and distortions in input data. The obtained results demonstrated that the developed method maintains high effectiveness and accuracy even under significant noise levels (up to 30 %), substantially expanding its practical application potential.
Fazulyanov, D.V. (fazulianov.dmitrii@gmail.com) - National Research Nuclear University “MEPhI” (Moscow Engineering Physics Institute) (Teaching Assistant); Guseva, A.I. (aiguseva@mephi.ru) - National Research Nuclear University “MEPhI” (Moscow Engineering Physics Institute) (Professor), Ph.D;
[№1 за 2026 год]
Authors: Antonov V.V. (antonov.v@bashkortostan.ru) - Ufa State Aviation Technical University, Faculty of IRT (Professor), Ph.D; E.V. Palchevsky (teelxp@inbox.ru) - Financial University under the Government of the Russian Federation; Sapozhnikov, A.Yu. (sapojnikovayu@umpo.ru) - Ufa Engine Industrial Association (Deputy Director), Ph.D; Mavrina, A.S. (nytka_008@mail.ru) - Ufa Engine Industrial Association (Head of Department, Leading System Administrator);
Abstract: Current trends in the creation and management of design documentation necessitate the continuous improvement of design and approval methodologies. The approach proposed in this work is based on the integration of a Large Language Model (LLM) with lean manufacturing principles. In this context, the digital environment of design engineering is treated as a space where traditional losses can be significantly reduced through the intelligent automation of design documentation control and approval. The developed LLM-based method for managing the electronic design documentation creation process automates key design stages: compliance checks against regulatory and technical standards, generation of standard review comments, and assignment of organizational units responsible for documentation approval. To demonstrate the efficacy of the proposed solutions, a web application was developed to integrate the methodology into existing business processes. As a result, the share of manual labor is substantially reduced, along with the number of revision cycles and the overall time required for electronic design documentation approval. The article presents a successful case study of the method's implementation, showing a 42.22 % reduction in approval time and a corresponding 57.78 % increase in process efficiency. Each technical control procedure is now performed faster than the normative benchmarks: model structure verification and comment generation take two minutes instead of the previous five. Thus, the neural network method based on LLM establishes a strong correlation between the requirements of design documentation and lean manufacturing principles. This yields a comprehensive effect: the approval cycle is accelerated, the volume of manual work is decreased, losses are reduced through automated comment generation, and the risk of errors is minimized. The transition to the new target state of the process reduced the number of approval iterations from twenty to six, demonstrating the practical value of the proposed method in the field of electronic design documentation management.
Keywords: neural network, LLM, 5S methodology, lean manufacturing, lean office, waste reduction, electronic design documentation
Visitors: 2673
Authors: Antonov V.V. (antonov.v@bashkortostan.ru) - Ufa State Aviation Technical University, Faculty of IRT (Professor), Ph.D; E.V. Palchevsky (teelxp@inbox.ru) - Financial University under the Government of the Russian Federation; Sapozhnikov, A.Yu. (sapojnikovayu@umpo.ru) - Ufa Engine Industrial Association (Deputy Director), Ph.D; Mavrina, A.S. (nytka_008@mail.ru) - Ufa Engine Industrial Association (Head of Department, Leading System Administrator);
Abstract: Current trends in the creation and management of design documentation necessitate the continuous improvement of design and approval methodologies. The approach proposed in this work is based on the integration of a Large Language Model (LLM) with lean manufacturing principles. In this context, the digital environment of design engineering is treated as a space where traditional losses can be significantly reduced through the intelligent automation of design documentation control and approval. The developed LLM-based method for managing the electronic design documentation creation process automates key design stages: compliance checks against regulatory and technical standards, generation of standard review comments, and assignment of organizational units responsible for documentation approval. To demonstrate the efficacy of the proposed solutions, a web application was developed to integrate the methodology into existing business processes. As a result, the share of manual labor is substantially reduced, along with the number of revision cycles and the overall time required for electronic design documentation approval. The article presents a successful case study of the method's implementation, showing a 42.22 % reduction in approval time and a corresponding 57.78 % increase in process efficiency. Each technical control procedure is now performed faster than the normative benchmarks: model structure verification and comment generation take two minutes instead of the previous five. Thus, the neural network method based on LLM establishes a strong correlation between the requirements of design documentation and lean manufacturing principles. This yields a comprehensive effect: the approval cycle is accelerated, the volume of manual work is decreased, losses are reduced through automated comment generation, and the risk of errors is minimized. The transition to the new target state of the process reduced the number of approval iterations from twenty to six, demonstrating the practical value of the proposed method in the field of electronic design documentation management.
Keywords: neural network, LLM, 5S methodology, lean manufacturing, lean office, waste reduction, electronic design documentation
Visitors: 2673
[№1 за 2026 год]
Visitors: 2673
Current trends in the creation and management of design documentation necessitate the continuous improvement of design and approval methodologies. The approach proposed in this work is based on the integration of a Large Language Model (LLM) with lean manufacturing principles. In this context, the digital environment of design engineering is treated as a space where traditional losses can be significantly reduced through the intelligent automation of design documentation control and approval. The developed LLM-based method for managing the electronic design documentation creation process automates key design stages: compliance checks against regulatory and technical standards, generation of standard review comments, and assignment of organizational units responsible for documentation approval. To demonstrate the efficacy of the proposed solutions, a web application was developed to integrate the methodology into existing business processes. As a result, the share of manual labor is substantially reduced, along with the number of revision cycles and the overall time required for electronic design documentation approval. The article presents a successful case study of the method's implementation, showing a 42.22 % reduction in approval time and a corresponding 57.78 % increase in process efficiency. Each technical control procedure is now performed faster than the normative benchmarks: model structure verification and comment generation take two minutes instead of the previous five. Thus, the neural network method based on LLM establishes a strong correlation between the requirements of design documentation and lean manufacturing principles. This yields a comprehensive effect: the approval cycle is accelerated, the volume of manual work is decreased, losses are reduced through automated comment generation, and the risk of errors is minimized. The transition to the new target state of the process reduced the number of approval iterations from twenty to six, demonstrating the practical value of the proposed method in the field of electronic design documentation management.
Antonov V.V. (antonov.v@bashkortostan.ru) - Ufa State Aviation Technical University, Faculty of IRT (Professor), Ph.D; E.V. Palchevsky (teelxp@inbox.ru) - Financial University under the Government of the Russian Federation; Sapozhnikov, A.Yu. (sapojnikovayu@umpo.ru) - Ufa Engine Industrial Association (Deputy Director), Ph.D; Mavrina, A.S. (nytka_008@mail.ru) - Ufa Engine Industrial Association (Head of Department, Leading System Administrator);
[№1 за 2026 год]
Authors: Sungatov, I.Z. (ilnazex@mail.ru) - Kazan National Research Technical University named after A.N. Tupolev–KAI (Associate Professor), Ph.D;
Abstract: This article examines mathematical and algorithmic approaches to constructing parametric models of helical surfaces for spherical milling cutters. Such tools are widely used in high-precision engineering for machining parts with complex spatial geometries. The work addresses theoretical and applied aspects of forming parametric models for helical surfaces characteristic of spherical cutters, which exhibit enhanced cutting properties. The primary focus is on integrating modern digital technologies into the tool design process. These include programming languages, numerical modeling methods, computer-aided design (CAD) technologies, and elements of artificial intelligence (AI). The methodological foundation of the research is a comprehensive approach encompassing mathematical modeling, algorithmic implementation of calculations in a Python en-vironment, construction of geometric profiles based on analytical expressions, and automated visualization of models in the KOMPAS-3D system. As a result, a system has been developed that ensures precise construction of the flute profile and automatic export of calculation data into a format compatible with the CAD environment. The created program significantly accelerates tool design, improves accuracy, and enables adaptation to production conditions. The algorithms are implemented as a software solution that constructs the 3D flute profile and automates the transfer of modeling results into the KOMPAS-3D engineering environment. The system also includes the capability to predict optimal tool parameters based on the analysis of previously designed models. The practical significance of the work lies in creating a digital platform capable of providing intelligent and automated cutting tool design, which is particularly relevant for manufacturing enterprises.
Keywords: tool design system, 3D object design, tool design, helical flutes, python, CAD, Kompass-3D
Visitors: 2718
Authors: Sungatov, I.Z. (ilnazex@mail.ru) - Kazan National Research Technical University named after A.N. Tupolev–KAI (Associate Professor), Ph.D;
Abstract: This article examines mathematical and algorithmic approaches to constructing parametric models of helical surfaces for spherical milling cutters. Such tools are widely used in high-precision engineering for machining parts with complex spatial geometries. The work addresses theoretical and applied aspects of forming parametric models for helical surfaces characteristic of spherical cutters, which exhibit enhanced cutting properties. The primary focus is on integrating modern digital technologies into the tool design process. These include programming languages, numerical modeling methods, computer-aided design (CAD) technologies, and elements of artificial intelligence (AI). The methodological foundation of the research is a comprehensive approach encompassing mathematical modeling, algorithmic implementation of calculations in a Python en-vironment, construction of geometric profiles based on analytical expressions, and automated visualization of models in the KOMPAS-3D system. As a result, a system has been developed that ensures precise construction of the flute profile and automatic export of calculation data into a format compatible with the CAD environment. The created program significantly accelerates tool design, improves accuracy, and enables adaptation to production conditions. The algorithms are implemented as a software solution that constructs the 3D flute profile and automates the transfer of modeling results into the KOMPAS-3D engineering environment. The system also includes the capability to predict optimal tool parameters based on the analysis of previously designed models. The practical significance of the work lies in creating a digital platform capable of providing intelligent and automated cutting tool design, which is particularly relevant for manufacturing enterprises.
Keywords: tool design system, 3D object design, tool design, helical flutes, python, CAD, Kompass-3D
Visitors: 2718
[№1 за 2026 год]
Visitors: 2718
This article examines mathematical and algorithmic approaches to constructing parametric models of helical surfaces for spherical milling cutters. Such tools are widely used in high-precision engineering for machining parts with complex spatial geometries. The work addresses theoretical and applied aspects of forming parametric models for helical surfaces characteristic of spherical cutters, which exhibit enhanced cutting properties. The primary focus is on integrating modern digital technologies into the tool design process. These include programming languages, numerical modeling methods, computer-aided design (CAD) technologies, and elements of artificial intelligence (AI). The methodological foundation of the research is a comprehensive approach encompassing mathematical modeling, algorithmic implementation of calculations in a Python en-vironment, construction of geometric profiles based on analytical expressions, and automated visualization of models in the KOMPAS-3D system. As a result, a system has been developed that ensures precise construction of the flute profile and automatic export of calculation data into a format compatible with the CAD environment. The created program significantly accelerates tool design, improves accuracy, and enables adaptation to production conditions. The algorithms are implemented as a software solution that constructs the 3D flute profile and automates the transfer of modeling results into the KOMPAS-3D engineering environment. The system also includes the capability to predict optimal tool parameters based on the analysis of previously designed models. The practical significance of the work lies in creating a digital platform capable of providing intelligent and automated cutting tool design, which is particularly relevant for manufacturing enterprises.
Sungatov, I.Z. (ilnazex@mail.ru) - Kazan National Research Technical University named after A.N. Tupolev–KAI (Associate Professor), Ph.D;
Keywords: tool design system, 3D object design, tool design, helical flutes, python, CAD, Kompass-3D
15. Repair crew task flow management based on priority rules
[№1 за 2026 год]Authors: Nasonov, M.A. (ma.nasonov@sintez-oka.ru) - LLC Sintez-OKA (Head of Department); Mantserov, S.A. (mca_9@nntu.ru) - R.E. Alekseev Nizhny Novgorod State Technical University, Institute of Industrial Technologies and Mechanical Engineering (Associate Professor, Director of Institute), Ph.D;
Abstract: This article addresses the problem of dynamic task flow management, illustrated through the example of repair crews at an industrial enterprise operating under changing equipment maintenance and repair schedules. As a methodological foundation, an algorithm for selecting scheduling strategies is proposed, based on workload management concepts and the application of simple priority rules. The mathematical model accounts for schedule dynamics, task urgency, crew workload levels, and task complexity. A key element of the model is an aggregated criterion enabling the adaptive selection of one of three minimization strategies: minimizing maximum tardiness; minimizing the number of overdue tasks considering time buffers; or minimizing the weighted sum of completion times. The results of computational simulation on a test dataset confirmed the effectiveness of the proposed approach. Utilizing the aggregated criterion reduces the number of overdue tasks. During simulation, on-time task completion ranged from 65–75 % compared to 60–70 % for fixed strategies, confirming the balance and adaptability of the proposed method and resulting in a more even workload distribution among crews compared to using fixed priority rules. The conducted analysis revealed potential for further algorithm development through the introduction of composite rules and automatic adjustment of weighting coefficients based on accumulated statistics. The obtained results can be used to build intelligent decision support systems for managing maintenance activities at industrial enterprises, contributing to improved resource utilization efficiency and enhanced reliability of production equipment.
Keywords: scheduling algorithms, method robustness, dynamic scheduling, workload distribution, load optimization
Visitors: 2643
15. Repair crew task flow management based on priority rules
[№1 за 2026 год]
Visitors: 2643
This article addresses the problem of dynamic task flow management, illustrated through the example of repair crews at an industrial enterprise operating under changing equipment maintenance and repair schedules. As a methodological foundation, an algorithm for selecting scheduling strategies is proposed, based on workload management concepts and the application of simple priority rules. The mathematical model accounts for schedule dynamics, task urgency, crew workload levels, and task complexity. A key element of the model is an aggregated criterion enabling the adaptive selection of one of three minimization strategies: minimizing maximum tardiness; minimizing the number of overdue tasks considering time buffers; or minimizing the weighted sum of completion times. The results of computational simulation on a test dataset confirmed the effectiveness of the proposed approach. Utilizing the aggregated criterion reduces the number of overdue tasks. During simulation, on-time task completion ranged from 65–75 % compared to 60–70 % for fixed strategies, confirming the balance and adaptability of the proposed method and resulting in a more even workload distribution among crews compared to using fixed priority rules. The conducted analysis revealed potential for further algorithm development through the introduction of composite rules and automatic adjustment of weighting coefficients based on accumulated statistics. The obtained results can be used to build intelligent decision support systems for managing maintenance activities at industrial enterprises, contributing to improved resource utilization efficiency and enhanced reliability of production equipment.
Nasonov, M.A. (ma.nasonov@sintez-oka.ru) - LLC Sintez-OKA (Head of Department); Mantserov, S.A. (mca_9@nntu.ru) - R.E. Alekseev Nizhny Novgorod State Technical University, Institute of Industrial Technologies and Mechanical Engineering (Associate Professor, Director of Institute), Ph.D;
