Інституційний репозитарій Вінницького національного технічного університету

Ласкаво просимо в інституційний репозитарій Вінницького національного технічного університету, який накопичує, зберігає, розповсюджує та забезпечує довготривалий, постійний та надійний доступ до наукових робіт професорсько-викладацького складу, співробітників, студентів, аспірантів і докторантів ВНТУ.

ISSN 2413-6360

Авторам

Фонди

Виберіть фонд, щоб переглянути його зібрання.

Нові надходження

Нелінійна динамічна модель освітнього процесу
Chornyi, О. Р.
;
Herasymenko, L. V.
;
Chorna, О. А.
(Вінницький національний технічний університет, 2026)
The objective of the article is to substantiate an approach to nonlinear modeling of the dynamics of the educational process as a controlled system learning outcomes are achieved through the combined action of a set of instructional stimuli and destabilizing factors of the educational environment, such as inertia, forgetting, and overload. The study employs methods of mathematical and simulation modeling, dynamic systems theory, local linearization, and numerical integration of differential equations. The controlled nonlinear “knowledge–disturbance” model has been proposed, in which the learning process is described through the interaction of two state variables: the normalized level of learning material acquisition and an integral indicator of the process intensity that reduce the efficiency of acquisition. This indicator generalizes the influence of external organization factors, cognitive overload, instability of attention, fatigue, stress, and other factors that are the cause of the uneven academic achievements over time. The paper shows that the classical linear second-order cybernetic model is suitable for describing inertia, damping, and effective losses in the educational process; however, it requires the extension in order to reproduce nonlinear saturation effects, background disturbances, and mode changes under improved learning conditions. For this purpose, the nonlinear controlled model is coordinated with a second-order cybernetic model by means of local linearization in the neighborhood of the equilibrium point. It is established that the parameters of the equivalent second-order equation can be interpreted through the derivatives of the right-hand sides of the nonlinear system, while the nature of transient processes is determined by the eigenvalues of the Jacobian matrix. Numerical modeling is carried out for a scenario involving a controlled change in learning conditions at a specified time instant. The results demonstrate the presence of three characteristic modes: initial adaptation, a quasi-stationary educational process with fluctuations, and transition to a new operating regime after the improvement of conditions. It is shown that reducing the intensity of destabilizing factors and increasing the efficiency of the instructional influence led to the increase in the average level of acquisition and decrease in process variability. The phase trajectory in the state space is proposed as an informative indicator of a change in the learning mode, since it reflects not only the average result but also the relationship between the achieved level of acquisition and the resource costs of the system. The scientific novelty of the study lies in the coordination of a controlled nonlinear model of the learning process with a second-order cybernetic model, which makes it possible to combine the pedagogical interpretation of acquisition, forgetting, and the influence of disturbances with the formal apparatus for analyzing stability and transient modes. The proposed approach can be used for simulation analysis of educational processes, assessment of the impact of organizational and methodological interventions, comparison of learning-load scenarios, and as a theoretical basis for developing adaptive and personalized learning systems.
Багатомовна класифікація електронних листів в умовах дисбалансу класів для маршрутизації в ректораті університету
Brusentsov, H. Ye.
(Вінницький національний технічний університет, 2026)
This study examines the issue of manual sorting of emails in the Rector’s office of a local university. Many emails received by the Rector’s secretariat require manual review and subsequent forwarding to other university departments. In this case, the emails were not processed directly by the secretariat; they had to be forwarded to seven different recipient departments. Emails intended for the departments of education, education quality, education outreach, the international office, science, the utility department and the vice-rector were analyzed. The initial email database contained 9.842 emails. After removing duplicates, 8.153 unique emails remained. Using an algorithm that combined duplicate removal with a search for similar emails yielded 4.672 unique records. Only 477 of these potential records could be considered sufficiently labelled for supervised model training. The three families of models investigated were: TF-IDF using logistic regression, TF-IDF using a support vector machine (SVM), and XLM-RoBERTa. The combination of TF-IDF and SVM yielded the best results. The combination of TF-IDF and SVM outperformed TF-IDF with logistic regression in terms of the F-measure (macro F1: 0.6421, weighted F1: 0.7920 and accuracy: 0.8021) on a random split and an F-measure (macro F1: 0.5371, weighted F1: 0.6854 and accuracy: 0.6947) on a chronological split. Class analysis demonstrated excellent results for the most common high-frequency classes, such as “science” and “education”. However, results for lower-frequency classes were variable. Recall for “education outreach” was zero for both splits due to the presence of only one test record. Of the 102 total errors made in both TF-IDF combinations, all but one occurred below a confidence threshold of 0.5. Thus, overall, this paper identifies the automatic forwarding of emails for the University Rector’s Office as a multilingual classification task with extreme class imbalance. The lack of sufficient labelled data and limited lexical models is the greatest obstacle to implement the solution for this task.
Методологія проєктування та дослідження інформаційно-вимірювальних систем витрати газу
Bilynskyi, Y. Y.
;
Stetsenko, A. A.
(Вінницький національний технічний університет, 2026)
The object of the study is the design process of ultrasonic information-measuring systems for determining gas flow rate and the ways to improve measurement accuracy under real operating conditions, which may be characterized by flow turbulence, geometric constraints of pipelines, and contamination of measuring channels. An iterative design methodology is proposed, which combines theoretical analysis of the problem, CFD modeling, device design development, experimental research, and artificial intelligence methods within a closed cycle “model – algorithm – hardware implementation – experiment – correction.” Artificial intelligence plays a significant role in the design process. The application of CFD modeling and intelligent data processing methods makes it possible to optimize the flowmeter design and significantly reduce the design time. The methodology is additionally enriched with local feedback loops, ensuring adaptability and correction at each stage of the design process. The use of such an iterative design model for ultrasonic flowmeters is possible with the availability of specialized equipment and software, which should include: an environment for numerical modeling of gas flows, a 3D printer, embedded computer systems for digital data processing, and test benches for studying the operation of measuring instruments. The application of the proposed methodology made it possible to: develop a flow-stabilizing device for the multi-beam ultrasonic gas meter GFA202 DN50, which allowed combining compact design with high accuracy typical of larger-size meters that meet class 0.5 requirements, although the meter itself belongs to accuracy class 1; create a new design of a four-channel flowmeter with equal signal paths, whose average measurement error does not exceed 0.8 % over a wide flow range, as the classical four-channel ultrasonic flowmeter has an error up to 1.3 %; propose a method for correcting gas flow measurement results based on the use of calibration data during calibration and verification of ultrasonic flowmeters. The implementation of the proposed solutions reduced the influence of hydrodynamic disturbances and structural factors on measurement results and improved the accuracy of the flowmeter. The obtained results can be applied in industrial gas transportation systems, particularly under conditions of limited lengths of straight pipeline sections and variable flow modes.
Діагностика стану лінійних динамічних систем методом роздільного оцінювання
Volovyk, A.Yu.
;
Savytskyi, А. Yu.
(Вінницький національний технічний університет, 2026)
To ensure high level of performance and operational reliability of modern, highly integrated technical systems, the timely detection and identification of malfunctions, failures, and sudden disturbances is critical. Their rapid identification makes it possible to promptly implement appropriate corrective actions, aimed at restoring system functionality and preventing the development of emergency situations. This paper addresses the problem of synthesizing a specialized filter designed to detect and localize failures, malfunctions, or disturbances in a class of linear stationary discrete-time systems. The fundamental principle of the algorithm is the concept of analytical redundancy based on a mathematical model of the system under observation. The diagnostic algorithm was designed so that pre-formed residual differences, in the form of signatures, would allow the effects of target-type faults to be distinguished the faults or disturbances influence of other classes. This resulted in the creation of an “extrapolator-corrector” device structure containing two parallel filters implemented in a form that is algorithmically equivalent to a Kalman filter. It is noteworthy that one of the subsystems provides a fault-free state vector estimate, while the other, which has a degenerate structure, is designed to generate a fault estimate. The resulting difference vector components are uniquely associated with a given fault type. The application of this algorithm reduces computational costs but introduces additional constraints and reduces the accuracy of the estimation results, due to computational issues related to the system’s pseudo-inverse matrices. Since this solution is approximate and non-unique, the resulting structure exhibits characteristics of quasi-optimality. In general terms, the convergence and stability of the obtained estimates are briefly discussed. The results of the method performance test are presented using an illustrative numerical example. The analysis of the results is presented in comparison with the abilities of the classical Kalman filter, which clearly demonstrates the effectiveness of the developed approach and its advantages. The computational experiment results confirmed the overall effectiveness of the method. The synthesized filter enables to generate state estimates of satisfactory quality and perform correctly when diagnosing unexpected malfunctions, failures, and sudden disturbances. Based on the conclusions drawn, it can be stated that this work introduces an element of scientific novelty into the field of fault detection and classification in linear discrete dynamic systems. Directions for further research have been identified.ілюстративному числовому прикладі. Аналіз отриманих результатів подано на фоні можливостей класичного фільтра Калмана, що дозволило наочно показати ефективність розробленого підходу та його переваги. Дані обчислювального експерименту підтвердили загальну працездатність методу. Синтезований фільтр здатний формувати оцінки станів задовільної якості та виконувати коректну роботу з діагностики непередбачених несправностей, відмов та різких збурень. На основі зроблених висновків можна констатувати, що представлена робота вносить елемент наукової новизни у сфері виявлення та класифікації множинних несправностей у лінійних дискретних динамічних системах. Сформовано напрямки подальших досліджень.
Синтез нечіткого регулятора для адаптивної системи безпеки колоборативного робота на основі комп’ютерного зору
Udovychenko, O.
;
Tsymbal, O.
(Вінницький національний технічний університет, 2026)
The article addresses the pressing scientific and technical problem of ensuring functional safety in conditions of close human-robot collaboration on modern automated lines. The research is conducted in the context of the global industrial transition to the Industry 5.0 paradigm, which requires the creation of flexible and human-centric production systems. The drawbacks of traditional safety safeguards (physical barriers, light curtains) are analyzed, the main of which is the rigid binary triggering logic. It initiates a complete emergency stop of the equipment upon any perimeter breach, leading to the interruption of technological cycles, increased downtime, and accelerated mechanical wear of the actuators. As an alternative, a method of dynamic Speed and Separation Monitoring (SSM) based on contactless computer vision is proposed. The authors have developed a three-level architecture of an intelligent safety system. At the first (sensor) level, the MediaPipe neural network framework is used, which processes the video stream in real time and generates spatial 3D coordinates of the operator’s skeleton joints. At the data processing level, a software module calculates the minimum distance between the human and the hazardous parts of the robot, and also differentiates its change to determine the kinematics of movement. The scientific novelty of the work lies in the synthesis of a Mamdani-type Fuzzy Controller for adaptive control of the manipulator’s speed, which, unlike existing solutions, implements a proactive strategy. Most known systems rely solely on static distance or actuator load (a reactive approach). The proposed system utilizes two linguistic variables: "Distance to the operator" and "Approach speed". This allows the controller to act preemptively: the system is capable of slowing down the robot in advance if a person approaches the zone with high acceleration. Such an approach ensures effective and safe dynamic scaling of virtual working zones (green, yellow, red). To verify the developed algorithm, mathematical modeling was performed in the MATLAB environment using the Fuzzy Logic Toolbox package. The application of the defuzzification procedure using the Center of Gravity method made it possible to obtain a smooth response surface. The results of the comparative analysis confirm that the synthesized fuzzy controller guarantees a stepless, smooth change in speed, completely eliminating the "chattering" effect of the control signal at the boundaries of virtual barriers. The practical significance of the work lies in the possibility of integrating the proposed architecture into the control systems of industrial cobots to optimize the balance between productivity and occupational safety.