MLMykhaylo Lobur

Mykhaylo Lobur

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mykhaylo Lobur has a total of 33 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Intellectual Information System for Semantic Analysis and Classification of Biomechanical Data

    by Mykhaylo Lobur, Dmytro Korpylov, Vira Oksentiuk, Zhanna Parashchyn

    The article proposes a method and intelligent information system for semantic analysis and automatic classification of human movement biomechanical data in rehabilitation systems. The method is based on the integration of a formal biomechanical description of movement with biomedical ontologies, which provides multi-level data processing: signal preprocessing, spatio-temporal feature extraction, ontological knowledge representation and automatic classification of movement patterns. A key feature of the developed system is the integration of a SPARQL engine for flexible ontology querying, SWRL rule verification and statistical analysis of rehabilitation data. A Python application using PyQt5, RDFlib, NetworkX and Matplotlib implements OWL ontology management, SWRL rule visualization, fuzzy membership functions and SPARQL query execution. The ontological model serves as a link between clinical movement analysis and decision support information systems, transforming semantic descriptions of movement into formalized knowledge for intelligent rehabilitation systems. The developed approach improves the interpretability of results, supports personalized therapy scenarios, and can be integrated into knowledge engineering systems for biomedical applications. Using SPARQL queries reduced data retrieval time by approximately 35-50% compared to traditional RDF traversal approaches, confirming the effectiveness of the proposed semantic processing architecture.

  • Using a time series databases as a component оf oil pumping units monitoring

    by Mykhaylo Lobur, Mykola Malyar

    The article considers the use of time series databases as part of SCADA systems for monitoring oil production facilities. The study focuses on the process of operational collection, storage and analysis of large volumes of technological parameters of the unit received in real time from sensors. The analysis revealed that traditional relational DBMSs are insufficient for processing high-frequency data streams at the engineering level. This complicates the prediction of failures and timely response to emergency situations. To achieve this goal, modern architectures and approaches to organizing time series storage were analyzed. This includes specialized file formats, hybrid solutions (TimescaleDB) and distributed systems (InfluxDB, QuestDB, GridDB). The criteria for selecting the optimal time series database for the engineering level of SCADA systems are proposed, taking into account performance requirements, ease of deployment, compatibility with industrial protocols, visualisation and implementation costs. A comparative analysis shows the ad-vantages of using InfluxDB, which provides write operations that are 5–10 times faster, flexible integration with Grafana, and support for open data exchange standards. Examples of practical application of InfluxDB for storing and visualising the parameters of an oil pumping unit are given. Unlike the classical time representation, the dynamogram method of constructing the dependencies of the installation parameters on each other is presented. The proposed approach to using InfluxDB for collecting and analysing parameters of the oil pumping unit at the engineering level of SCADA systems can be applied to the design and operation of operational control and monitoring systems of oil fielda.

  • Open Access

    ANALYSIS OF THE STATUS AND PROBLEMS OF THE OPERATION OF WIRELESS AD HOC AND MESH NETWORKS

    by Oleksii Dzhus, Mykhaylo Lobur

    The article presents a comprehensive analysis of the state and problems of wireless ad hoc and mesh networks, with an emphasis on their architectural features, classification, application areas, and key challenges. Wireless ad hoc networks, such as MANET, VANET, WSN, FANET, UWVANET, SANET, and HANET, are characterized by self-organization, multi-hop routing, and high mobility of nodes, which makes them ideal for scenarios without fixed infrastructure, such as military operations, disaster response, intelligent transportation, and environmental monitoring. At the same time, mesh networks provide a static backbone with gateways, multi-radio interfaces, and integration with other networks, offering improved throughput, stability, and scalability, but necessitating careful resource management. The analysis of recent research encompasses the architectures of wireless mesh networks, routing protocols (such as HWMP, OLSR, and BATMAN), multi-channel designs, and hierarchical models. The research methodology is based on theoretical models, empirical analysis of problems (including interference, energy constraints, and topology instability), and comparative analysis. The results demonstrate that classical theory shows a drop in throughput in large networks due to interference and channel sharing. At the same time, the delay accumulates as the sum of hop times, taking into account queuing and processing. Special attention is paid to the role of intermediate nodes in multi-hop networks, as they provide routing, load balancing, resilience, and coverage extension, but also create challenges such as overhead and risks of malicious behavior. Efficiency issues include interference, energy trade-offs, deployment complexity, and technological limitations (e.g., LoRa duty cycle). The bandwidth and delay constraints are analyzed using Shannon's and cumulative delay formulas, with examples illustrating degradation in mobile scenarios. Topology instability due to node mobility is modeled by the probability of link breakage, and energy constraints are modeled by the energy consumption per transmission.

  • MODELING OF THERMOELECTRIC PROCESSES IN A FAULTY LITHIUM-ION BATTERY WITH A LOST CELL

    by Lviv Polytechnic National University, Oleksandr Melnychuk, Mykhaylo Lobur, Vasyl Ivanyna

    Wide range of malfunctions can occur during the operation of lithium-ion battery packs (BPs). However, most existing studies focus on internal short circuits and thermal runaway of the battery detection. The same statement also applies to the most common monitoring methods of BP operating parameters and their faults detection. Therefore, such methods may be incapable of identifying a malfunction such as degradation or a complete loss of electrical connection to a cell within the battery. At the same time, losing a connection to even a single cell in the BP can lead to significant changes in operating modes, affecting the overall safety, efficiency, and longevity of the battery. Modeling this type of failure is an essential step in developing new and more effective diagnostic methods for BPs. This study models a 3s7p battery pack consisting of lithium-ion 18650 cells with a lost connection to one of the cells. Numerical modelling is conducted utilizing COMSOL Multiphysics software. A mathematical model of the thermoelectric processes in the battery is described. The finite element method is used to calculate the thermal field during battery discharge. The impact of the lost cell on neighbouring cells and the overall BP is examined. The modeling results confirm that the loss of a cell creates an imbalance in current distribution among the battery cells, leading to accelerated heating and rapid discharge of the cells connected in parallel with the lost one. Although the remaining cells discharged more slowly, they cannot compensate for the overall loss of battery state of charge. These changes increase the risk of overheating the functional cells and may contribute to their rapid degradation in the future. The thermal field inside the battery is calculated. Its analysis shows distinct differences when compared to a healthy battery. The results may be useful for further development of improved BP design and thermal management methods. The recommended method can be utilized for diagnostics and early fault detection in lithium-ion battery packs, thereby decreasing operational risks and enhancing battery longevity.

  • THE INTEGRATION OF ELECTROENCEPHALOGRAPHY AND VIRTUAL REALITY FOR REHABILITATION: A PERSPECTIVE REVIEW OF SUCCESSES AND PITFALLS

    by Lviv Polytechnic National University, Oleksii Dzhus, Mykhaylo Lobur

    The combination of neurotechnology and advanced digital environments has opened promising opportunities in rehabilitation. Integrating electroencephalography (EEG) with virtual reality (VR) offers a novel approach to remedy by combining real-time neural monitoring with immersive, interactive environments. This perspective review examines the successes and pitfalls associated with EEG-VR applications in rehabilitation, highlighting how these modern technologies can enhance patient outcomes. The other part of this review identifies current challenges that delay the widespread clinical adoption of this technology combination. Methods: A comprehensive literature review was conducted across PubMed, Scopus, and Web of Science databases from their inception to February 2025. We identified the studies investigating the combined use of EEG and VR in rehabilitation and reviewed them using the SANRA method. Each study was then evaluated based on its methodology, patient demographics, technological integration strategies, therapeutic outcomes, and reported limitations. Results: Out of 105 articles, 65 remained after filtering. 11 articles were included to our analysis after the title screening, as well as abstract and full-text reviews. The review identified growing evidence supporting the efficacy of EEG-VR systems in motor recovery, cognitive function, and neurofeedback training across various patient populations. We found EEG-VR therapy successes in enhanced user engagement and the ability to adapt rehabilitation protocols based on real-time neural data. On the other hand, the analysis also has shown pitfalls, including technical challenges such as signal interference, synchronization issues between EEG and VR platforms, and a lack of standardized protocols, limiting outcomes' scalability and consistency. Conclusions: The integration of EEG and VR is an emerging area in rehabilitation, offering innovative approaches for personalized and adaptive therapy. Despite encouraging preliminary results, current technological and methodological hurdles need further investigation. Future research should focus on optimizing system integration, standardizing intervention protocols, and conducting large-scale clinical trials to investigate thoroughly the usage of EEG-VR systems in rehabilitation.


  • IMPROVEMENT OF THE POSITIONING SYSTEM OF ROBOTIC DEVICES BASED ON THE COLOR BEACON ALGORITHM

    by Lviv Polytechnic National University, Oleksandr Belej, Oleksii Dzhus, Mykhaylo Lobur, Danylo Golovanchuk

    Today, the development of indoor positioning systems is a promising and relevant, but difficult task. This requires the creation of maps based on floor plans, the selection of effective location algorithms, and the creation of an appropriate building infrastructure for reliable location determination. Existing positioning systems that use radio channels for indoor navigation, infrared and ultrasonic tagging are considered. It is established that the use of a radio channel indoors is problematic due to interference and signal reflections that affect accuracy. The analysis showed that the most reliable channel for use indoors is a channel in the visible range of the electromagnetic spectrum. Such systems work using video cameras and algorithms. Systems known from open sources that use video cameras and machine vision systems are considered. The main problem of existing indoor positioning technologies based on applied video surveillance systems is their connection with the working background, such as the interior, or the requirement of a uniform background for the use of colored beacons. Based on the analysis, the use of colored beacons as reference points and a video measurement system for calculating distances to these points for indoor navigation are proposed. The use of colored beacons is proposed based on the hypothesis that they can work in conditions of a heterogeneous background, unlike existing systems. To confirm these assumptions, appropriate studies were conducted. An algorithm for recognizing colored lights on an input video image obtained from a digital image processing video camera has been developed. The algorithm is based on constructing a "color mask" using a smooth continuous logistic sigmoid curve and a Gaussian function to extract the color characteristic of the light signal. An algorithm for determining the coordinates of the marker recognition point based on recognized colored beacons in frames of the input video image has been developed. To determine the coordinates of a mobile robotic system, measurements are taken on the video image to determine the relative coordinates of the beacons in the camera coordinate system. A relative map is constructed, which is converted into absolute camera coordinates in the world coordinate system using a three-dimensional transformation. This is made possible by the predefined absolute coordinates of the color beacons. Filtering parameters for the color beacon recognition algorithm are defined, which allows the algorithm to maintain performance under conditions of heterogeneous backgrounds. The following parameters are used: the variance of the Gaussian curve for hue filtering, the curve and logistic threshold shift of the sigmoid curve for saturation; the logistic curvature of the sigmoid curve and the brightness threshold shift. These parameters allow classifying the pixels of the input image by color, thereby identifying only those areas that correspond to the color values of the color beacon areas under variable lighting conditions.

  • Open Access

    The Pushing Mechanism Design of Jumping Robot

    by Roman Zinko, Mykhaylo Lobur, Andriy Zdobytskyi, Tetyana Stefanovych

    . In areas hazardous for humans, jumping robots serve as substitutes. They exhibit enhanced mobility when traversing uneven or rugged terrain, significantly broadening the application potential of mobile devices for territory surveillance, environmental monitoring after earthquakes, and other emergency situations where robot movement relies solely on jumping. Various design solutions for performing jumps are known, each with its own advantages, disadvantages, and implementation based on specific structural elements. To propel robots off the ground, different actuators can be used: spring-based mechanisms (with or without additional kinematic links), pneumatic or hydraulic systems, and those powered by fuel-air mixtures. Spring-based impact systems execute jumps by converting the potential energy stored in the spring into the kinetic energy of the robot's body. The objective of the research is to justify the feasibility of developing cam mechanisms for jumping robots, taking into account the direction of their jumps in accordance with the distribution of forces and applied loads. The methodology involved determining the impact force of the hammer in the cam mechanism under static loading. A calculation method for the parameters of the impact system was developed. Computer modeling of the spring compression phase using a lever device was conducted. The structural and technological parameters of the robot were substantiated, considering its mass, spring stiffness, impact system, and the cam profile, using Autodesk Inventor software. Calculations of forces, power, and design parameters of the impact system were presented, along with a detailed analysis of all phases of the jumping process. These phases include spring compression using a lever device, removal of the compression device, spring release, and the final ballistic phase (flight phase). The methodology allowed for determining the design parameters of the cam mechanism based on specified kinematic and dynamic parameters of the jumping process.

  • Open Access

    Automation of the Oil Extraction Process Performed by Means of A Screw Press

    by Mykhailo Lobur, Vitaliy Korendiy, Oleksandr Kachur, Volodymyr Havran

    The continuous development of the oil-manufacturing industries causes the necessity of improving extraction technologies. In this case, the specific interest is focused on the control systems of screw presses. Among a great variety of such machines, the small household presses are in significant demand among consumers. Various seeds and kernels require different technological conditions to be provided in order to maximize the qualitative and quantitative characteristics of the extracted oil. Therefore, the main objective of this research is developing and testing the control system allowing for automation of the oil extraction process. Particularly, the temperature parameters of the pressing chamber, extracted oil, and electric motor are to be monitored and limited. In addition, the consumer should be able to predefine the mass of the oil to be extracted. Considering the small household screw press LiangTai LTP200, the general algorithm (block diagram) of the control system operation is proposed and the corresponding experimental prototype is developed. The latter is based on the Arduino Mega microcontroller and is equipped with three temperature sensors, two coolers (fans), one heater, and one mass sensor. The proposed control system allows for continuous monitoring and limiting of the pressing chamber, oil, and electric motor temperatures, as well as the mass of the extracted oil. The experimental data show that the pressing chamber preheating process lasts for about 3 min (170…190 s) and its maximal temperature does not exceed 44°C. The temperature of the extracted oil does not rise over 61°C. The motor temperature changes within the range of 69...71°C. The oil extraction productivity is as follows: 1.2 kg/h (sunflower seeds), 1.06 kg/h (walnut kernels), 0.9 kg/h (almond kernels), and 0.78 kg/h (peanut kernels). The obtained results can used in further investigations focused on analyzing the influence of these parameters on the quantitative and qualitative characteristics of the extracted oil.


  • Open Access

    PROBLEMS OF MACHINE LEARNING IMPLEMENTATION ON MODERN EDGE DEVICES

    by O. S. Melnychuk, M. V. Lobur, T. V. Sviridova

    Recently, artificial intelligence (AI) and machine learning (ML) have attracted considerable attention in both industry and academia. Since traditional ML methods are energy-demanding, they can be applied only to a limited subset of devices with significant computing capabilities. However, the rise in complexity of available processors and microcontrollers made it possible to integrate ML methods into data pipelines on mobile devices, and later – on low-power (few milliwatts) edge devices. This concept finally took shape in 2018 and was named TinyML. Its rapid spread is fueled by the huge number of manufactured microcontrollers (250 billion) and the popularity of the Internet of Things (IoT). In the future, TinyML can be used on every device that is part of the IoT, although this will not always be economically justified. At the same time, the application of TinyML technologies is not straightforward and requires a careful assessment of the capabilities as well as selection of both the ML algorithm and the peripheral device. This is explained primarily by the lack of generally accepted benchmarks of both TinyML algorithms and hardware capabilities for the application of ML. There are several approaches to TinyML implementation: softwareoriented, hardware-oriented, and hybrid. Depending on the approach, different methods of solving a specific problem are used. Some ML algorithms can be adopted to edge devices through the use of simplification and adaptation techniques tailored for neural networks (NN). However, the process is not automated and generalized as of now. This article is devoted to an overview of the concept of TinyML, its main stages and features, the most prominent achievements in the fields of speech and image recognition, sequence classification and data compression, health diagnostics and brain interaction, prediction of equipment malfunctions and anomaly detection, autonomous transport and ecology. Unfortunately, the volume of the article does not allow us to reveal the specifics of TinyML’s application in other fields and with more details. In addition, the article analyzes the problems that arise during ML implementation on low-power devices. The purpose of the article is to become a short guide and roadmap to the world of TinyML applications. Останнім часом штучний інтелект (ШІ) та машинне навчання (МН) привернули значну увагу як на виробництві, так і в академічних колах. Оскільки традиційні методи МН є енергоємними, то це обмежує їх застосування підмножиною пристроїв зі значними обчислювальними можливостями. Однак зростання потужності процесорів та мікроконтролерів дозволило інтегрувати методи МН для обробки даних спершу на мобільні пристрої, а пізніше – на малопотужні (кілька міліват) периферійні пристрої. Ця концепція остаточно сформувалась у 2018 році та отримала назву TinyML. Її стрімке поширення підживлюється велетенською кількістю вироблених мікроконтролерів (250 мільярдів) та популярністю інтернету речей (ІР). У перспективі, TinyML може бути використаний на кожному пристрої що є частиною ІР, хоча це не завжди буде економічно виправдано. Водночас застосування технологій TinyML не є простим і потребує ретельної оцінки та вибору можливостей як алгоритму МН так і периферійного пристрою. Це пояснюється насамперед відсутністю загальноприйнятих способів порівняння як алгоритмів TinyML, так і спроможностей апаратного забезпечення щодо застосування МН. Існує кілька підходів до впровадження TinyML: орієнтований на програмне забезпечення, апаратне забезпечення та гібридний. Залежно від підходу застосовуються різні методи вирішення конкретної задачі. Деякі алгоритми МН можуть бути перенесені на периферійні пристрої досить просто, через застосування прийомів спрощення та адаптації нейронних мереж (НМ). Однак на сьогодні їх застосування не є автоматизованим та узагальненим. Ця стаття присвячена огляду концепції TinyML, основних етапів та особливостей її застосування, найбільш помітних досягнень в галузях розпізнавання мови та зображень, класифікації послідовностей та стиснення даних, діагностики здоров’я та взаємодії з мозком, прогнозування несправностей техніки та виявлення аномалій, автономного транспорту та екології. На жаль, об’єм статті не дозволяє розкрити особливості застосування TinyML в решті галузей та більш...