AEEE. 2026, vol. 24

Permanent URI for this collectionhttp://hdl.handle.net/10084/158428

Browse

Now showing 1 - 15 out of 15 results

EBMDP: Equal Balance Message Drop Policy for QoS Optimization in Delay Tolerant Networks
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Ahmad, Zahoor; Saeed, Khalid; Anwar, Muhammad Shahid; Frnda, Jaroslav; Fajeed, Muhammad Faran; Chromý, Erik; Khan, Samiullah; Kutlimuratov, Alpamis
elay-tolerant networks (DTNs) are ad hoc in nature. It is known for characteristics such as in- termittent connectivity and dynamic topology. How- ever, a unique characteristic of DTNs is that there is no guarantee of an end-to-end connection between sender and receiver. Therefore, nodes observe long de- lays in establishing the connection. Once a connec- tion is established between the nodes, the links connect- ing the nodes are fully utilized, and the buffer mem- ory overflows, resulting in congestion that significantly compromises the quality of services (QoS). To avoid congestion, researchers have developed different buffer management policies. This research presents an effi- cient buffer management policy, known as the Equal Balance Message Drop Policy (EBMDP), designed to improve QoS in DTN. The EBMDP discourages un- necessary message drop. EBMDP drops selected mes- sages from the overflowed node, and the selection of messages for the drop from the overflowed node is based on the conditions defined by the EBMDP. The results of EBMDP are better than the drop-oldest ap- proach (DOA) and size-aware drop (SAD) regarding delivery probability (DP), overhead ratio (OR), buffer time average (BTA), and dropped messages. The de- livery probability of EBMDP obtained by simulation is 0.1861, which is higher than the delivery probabilities of SAD and DOA, which are 0.1069 and 0.1114, re- spectively. Similarly, the overhead ratio of EBMDP is lower than that of SAD and DOA. The results show a significant improvement in the buffer time average, as the buffer time average of messages using EBMDP is greater than that of SAD and DOA. The results also show lower messages dropped (MD) for EBMDP than for MD of SAD and DOA.
Bio-Inspired Optimization and Machine Learning for Multi-Band Impedance Matching Networks
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Amuda, Abdulrasaq Olanrewaju; Karataev, Tologon; Oshiga, Omotayo; Osanaiye, Opeyemi; Stittu, Moshood; Obetta, James; Araoye, Timothy Oluwaseu
The intelligent design of multi-band impedance matching networks was investigated through the integration of bio-inspired optimization and ma- chine learning classifiers. The Hippopotamus Opti- mization Algorithm (HOA) was employed in conjunc- tion with Support Vector Machine (SVM), Artificial Neural Network (ANN), and Random Forest models to derive accurate and fabrication-ready design param- eters. The optimal configuration, defined by a width of 2.7936 mm, spacing of 0.6103 mm, and length of 1.0893 mm, produced a reflection coefficient (S11) of −29.1456 dB, indicating excellent impedance matching across the target frequency band. Among the classi- fiers, the SVM achieved the highest generalization ac- curacy of 96.76% and the lowest mean squared error of 0.3174, surpassing the performance of ANN and Ran- dom Forest. The developed framework reduces reliance on computationally intensive electromagnetic simula- tions, shortens design time, and maintains high predic- tive precision. These results confirm the effectiveness of combining evolutionary optimization with machine learning for the efficient and compact design of multi- band RF matching networks.
HemoGAT: Heterogeneous multimodal speech emotion recognition with cross-modal transformer and graph attention network
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Nguyen, Nhut Min; Nguyen, Thanh Trung; Nguyen, Tien-Dat; Dang, Duc Ngoc Minh
Multimodal speech emotion recognition (SER) is a promising field, yet effectively fusing diverse information streams remains challenging. Addressing this requires architectures capable of modeling structural relationships across modalities with fine-grained, feature- level interactions. This paper proposes HemoGAT, a novel heterogeneous multimodal SER architecture that integrates a dual-stream architecture with two core mod- ules: a heterogeneous multimodal graph attention net- work (HM-GAT) and a cross-modal transformer (CMT) to address this. The HM-GAT module captures complex structural and contextual dependencies using a hetero- geneous graph constructed from deep embeddings. The CMT module enables precise cross-modal feature fusion through bidirectional cross-attention. This design effec- tively captures both high-level relationships and immedi- ate cross-modal influences. HemoGAT achieves state-of- the-art (SOTA) performance on the IEMOCAP dataset and highly competitive results on the MELD dataset, demonstrating its superiority over existing methods. Extensive ablation studies were conducted to evaluate HemoGAT. We assessed the impact of the Top-K algo- rithm for heterogeneous graph construction and com- pared unimodal and multimodal fusion strategies. We also examined the contributions of the HM-GAT and CMT modules, analyzed the role of the graph attention network (GAT) in graph learning, and evaluated the effect of GAT layer depth on performance
Features Of Boundary Condition Formation For The Long Transmission Line Equation Using Equivalent Circuit Approaches
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Levoniuk, Vitaliy; Lysiak, Heorhii; Lysiak, Vladyslav
This paper presents a comparative analysis of the application of Γ- and Π-type equivalent circuits for modeling boundary conditions to the long transmis- sion line equation with distributed parameters. The main objective is to assess how the choice of substi- tution scheme for the first and last discrete segments of the line affects the accuracy of transient process sim- ulation and the symmetry of voltage and current dis- tribution along the line. Mathematical models of an intersystem overhead AC transmission line in a single- phase representation were developed, the wave equation was discretized using the method of lines, and the simu- lations were implemented in the Fortran programming environment. Two numerical experiments were con- ducted: the first using a direct Γ-type equivalent cir- cuit, and the second using a symmetric Π-type equiv- alent circuit. In each experiment, the line was ener- gized alternately from both ends, enabling the evalua- tion of model symmetry. The simulation results show that the Π-type scheme ensures complete symmetry of the electromagnetic quantities regardless of the direc- tion of power flow, in contrast to the Γ-type scheme, which introduces minor asymmetry. The obtained find- ings justify the choice of equivalent circuit depending on the required model accuracy and complexity, and may be useful in the development of modern methods for analyzing non-stationary regimes in high-voltage trans- mission networks.
Development of internet of things based flood monitoring system with real-time dashboard at flood monitoring project seelab kencana SDN. BHD. Shah Alam
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Bagaskara, Alvino Tanjung; Irawati, Indrarini Dyah; Mahzan, Muhammad Akmal Bin; Bakar, Muhamad Husaini Bin Abu
This research develops an Internet of Things (IoT)-based flood monitoring system equipped with a real-time web dashboard and artificial intelli- gence (AI)-based prediction features, as a solution to the frequent flooding in Shah Alam, Malaysia, due to high rainfall and a sub-optimal drainage system. The system uses AJ-SR04M sensors to measure water lev- els, MH-RD sensors to detect rainfall, and ESP32 and ESP32-CAM for data processing and image cap- ture. Data is sent wirelessly to a Django-based back- end server and displayed on a web dashboard. The backend also processes the data and runs a time-series forecasting-based machine learning model to predict conditions five minutes ahead, with the predicted re- sults displayed alongside the actual data. In addition, the system provides automatic notifications via Tele- gram when sensor values exceed a threshold. The test results show that the system is able to display envi- ronmental data accurately and responsively, provide real-time early warnings, and generate predictions that match historical trends. The system has successfully supported effective flood risk mitigation in vulnerable areas by providing accurate sensor data and AI-based predictions that match historical trends.
Comparison of dual three-phase and six-phase surface maunted permanent magnet motor for traction drive applications
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Quy, Duong Le Doan; Nguyen, Duc Quang; Chi, Phi Do; Luong, Tuan Nghia; Quoc, Vuong Dan
Permanent magnet synchronous motors (PMSMs) are widely used in industrial and commer- cial applications such as electric vehicles, robotics, and aerospace, owing to their high power and torque density, operational stability, and exceptional efficiency. The integration of permanent magnets (PMs) eliminates the need for an external excitation current, thereby minimiz- ing excitation losses. Recently, six-phase motor systems have emerged as viable alternatives to traditional three- phase configurations, offering enhanced fault tolerance and improved control performance under fault condi- tions. Critical research areas for PMSMs in electric ve- hicle applications include the design and implementation of advanced control strategies to achieve high precision, reliability, and energy efficiency. This study proposes a comparative electromagnetic performance analysis of dual three-phase and six-phase surface-mounted per- manent magnet motors, utilizing both analytical and finite element methods. To further optimize the pro- posed motors, a segmented skewing technique is applied, demonstrating additional improvements in performance characteristics. This research contributes to a deeper understanding of these motor configurations for traction drive applications.
Enhancing Scalability of LoRa Networks in Smart City Conditions: A Machine Learning Approach
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Malik, Mandee; Kothari, Ashwin; Pandhare, Rashmi
The Internet of Things (IoT) is a rapidly expanding network of interconnected devices. Low Power Wide Area Network (LPWAN) technologies un- der the umbrella of IoT enable cost-effective long-range communication. Among LPWAN technologies, LoRa has gained prominence as a leading unlicensed-band IoT technology, particularly suited for smart city de- ployments due to its ability to transmit over several kilometers with minimal power consumption and low data rates. However, in densely populated urban set- tings, LoRa devices transmit independently, often lead- ing to packet collisions and degraded network perfor- mance. This study introduces an optimized approach to configuring LoRa communication parameters to im- prove scalability, reduce collisions, and enhance data transmission reliability. We propose a stochastic gradi- ent descent (SGD)-based optimization method, achiev- ing a 2–12 % increase in delivery ratio across net- works ranging from 200 to 2000 devices. Additionally, we evaluate and compare existing optimization strate- gies, including MinSF, ADR, and ADR+, while refin- ing Spreading Factor (SF) allocation to mitigate inter- ference. The findings highlight a significant enhance- ment in LoRa network efficiency, making it more reli- able for large-scale smart city applications.
A Study of Ensemble Models for Defect Prediction from Class Diagram
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Battulga, Batnyam; Tsoodol, Lkhamrolom; Erdenebaatar, Bilguu; Erdenebaatar, Tsetsegjarg; Namsrai, Oyun-Erdene; Bold, Naranchimeg
Software defect prediction in the early stages of the Software Development Life Cycle (SDLC) is crucial to reducing project cost and ensuring the implementation’s success. Existing methods for software defect detection in a project rely on the implementation or testing phases of the SDLC, based on the source code. While relatively few studies have focused on identifying defects in the design phase of the SDLC, these approaches primarily employ machine learning or deep learning methods to detect and classify suspect code segments or classes in static diagrams as defective or clean. This study utilizes 24 model-based metrics extracted via SDMetrics, including structural and objectoriented design features derived from UML class diagrams. To enhance classification performance, this study introduces an ensemble machine learning model with different techniques (stacking, voting) that combine multiple machine learning models. Specifically, we compare ensemble models with different ensemble techniques to the individual models in terms of accuracy, precision, recall, F-measure, and AUC by utilizing a large dataset called the Unified Bug Dataset, comprising five publicly available sub-datasets. Experimental results show that the ensemble model with the stacking ensemble method outperformed other ensemble models and the individual classifiers (RF, XGBoost, ET) in terms of AUC.
Fractional order proportional integral dertivative dased swarm optimization for three-tank system
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Dulaimi, Hussien; Al-Khazraji, Huthaifa; K. Hamzah, Mohammed
The control of a liquid level in the most of the industrial process gains high importance, par- ticularly in petrochemical, food processing and phar- maceutical. This paper deals with the designing of a fractional-order proportional-integral-derivative (FOPID) technique to managing liquid level in a three- tank system. To get the best performance from the FOPID controller, its adjustable parameters were fine- tuned using the crow search algorithm (CSA) and mine blast algorithm (MBA). The effectiveness of the pro- posed optimized FOPID controller was further exam- ined with standalone PID controller based on computer simulation using MATALB. Based on results of the time-based metrics like overshoot and how quickly the system settles, the FOPID-CSA controller outperforms the FOPID-MBA, the PID-CSA and the PID-MBA controllers. Moreover, the results obtained from robust- ness analysis showed that the FOPID-CSA controller is more robust under disturbances.
Particle swarm optimization in the field control of a novel electric vehicle design based on a linear induction motor
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Berrahal, Sebti; Chikhi, Abdesslem; Khettache, Laid
This work aims to improve the performance of electric vehicles (EVs) based on linear induction mo- tors (LIM). The Particle Swarm Optimization (PSO) method is proposed to tune the PID regulator of the Field-Oriented Control (FOC) technique. The main objective of this study is to develop innovative solutions that maximize the efficiency and precision of electric vehicles on various paths. The LIM model is imple- mented using the d-q synchronous reference frame and takes into account the end-effect phenomenon. This phenomenon occurs due to the termination of the mo- tor’s physical structure, which leads to distortion in the magnetic field at the ends of the motor’s primary (sta- tor). It is also highly nonlinear, which increases its complexity and makes control difficult. To overcome this issue, the Field-Oriented Control (FOC) technique is suggested to achieve better efficiency, dynamic per- formance, and greater control flexibility of the motor. Furthermore, the use of the (PSO) optimization tech- nique enables the determination of optimal control pa- rameters to maximize the performance of the (FOC- LIM) system under different operating conditions, such as speed variation and disturbance load. A compari- son between the PSO-PID and conventional methods in terms of response stability, steady-state error, and rise time is conducted using MATLAB/Simulink. The results demonstrate a more efficient, precise, and high- performing electric vehicle system.
Performance analysis of energy harvesting–enabled lora networks under hardware impirments with diversity and learning techniques
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Nguyen, Thi Tuyet-Hai; Nguyen, Hong Son; Huynh, Trong Thua; Phan, Nghia Hiep
This paper investigates the performance of energy harvesting (EH)–enabled long range (LoRa) networks with diversity techniques under the impact of hardware impairments (HI). In particular, we analyze the coverage probability (Pcov) of a network where LoRa end devices (EDs) rely solely on harvested energy supplied by power beacons (PBs). Both the gateway and PBs are equipped with multiple antennas and employ maximal ratio combining (MRC) and maximal ratio transmission (MRT) techniques, respectively, to enhance system performance. Due to the complexity of the considered network, conventional mathematical analysis becomes intractable. To overcome this challenge, we leverage machine learning and deep learning approaches including support vector machines (SVMs), random forest (RF), gradient boosting (GB), and neural networks (NNs) with different normalization strategies to estimate the coverage probability of the system. Extensive simulation results show that neural networks provide the most accurate performance predictions, followed by SVMs and RF, while GB exhibits the weakest performance. Nonetheless, the performance gaps among these models remain moderate to minor, indicating that all are suitable for most Internet of Things (IoT) applications. The accompanying parameter sensitivity analysis highlights the critical roles of the path-loss exponent and PB transmit power, offering valuable design and optimization insights for EHenabled LoRa networks under practical hardware constraints.
Bacteria in blood identification using electronic nose data based on LSTM and BILSTM deep neural network models
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Sedhane, Mouna; Hafs, Toufik; Daas, Sara; Hatem, Hatem
Bacteria are single-celled organisms that en- ter the body, grow, and release toxins that harm cells, causing sepsis and other diseases. Because bacteria cause various diseases in humans, prompt diagnosis is required to adapt antibiotic medication and prevent disease spread. This study presents a promising de- vice that can distinguish between different types of bac- teria commonly found in the blood. Electronic nose technology is now regarded as a quick tool for detect- ing pathologies based on volatile organic compounds (VOCs). The use of classical bacteriology takes time to give the practitioner or biologist a diagnosis. The bacterial species is detected from VOCs released by bac- teria in a few minutes using a multi-sensor system for the detection of VOCs. The goal of this study was to test and identify ten different types of bacteria in blood by an electronic nose. The proposed models achieved accuracies of 96.77% (LSTM) and 98.91% (Bi-LSTM), demonstrating the superiority of Bi-LSTM for bacterial classification.
Lightweight deep learning for autonomous human counting system on low-cost hardware
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Le, Anh Vu; Le, Nhat Tan; Nguzen, Anh Dung; Nguzen, Ngoc Nghia; Le, Hai Dang; Tran, Minh Dang; Minh, Bui Vu; Huynh, Lam Dong; Elara, Mohan Rajes
Accurate and efficient human counting is essen- tial for optimizing public transportation and advancing smart city infrastructure. This paper evaluates pro- posed lightweight deep learning models for autonomous human counting system on low-cost hardware, ensur- ing real-time monitoring and enhanced operational ef- ficiency. While existing methods, such as DeepSORT, Kalman Filters, and YOLO variants, are often im- plemented on high-end hardware, they typically prior- itize accuracy over computational efficiency. Few ob- ject detection and tracking techniques can run in real- time on low-end hardware. This work advances the field by utilizing optimized deep learning models suit- able for embedded systems with constrained resources. Specifically, fine-tuned YOLOv8 is employed for head detection, combined with ByteTrack for robust track- ing, outperforming YOLOv5 and YOLOv11 in accu- racy and efficiency. Archiving the 15 FPS and more then 90% accuracy on the real environment deployment on both RISC-V architecture with an integrated NPU (MaixCAM) and ARM v8 (Raspberry Pi), The pro- posed system demonstrates its suitability for real-time, cost-effective, and scalable autonomous human count- ing in public transit environments.
Resilient control scheme for LVRT enhancement in DFIG system with thyristor-contolled LC compensator
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Roy, Anushree; Debnath, Sudipta
This paper presents a multi-functional thyristor-controlled LC (TCLC) compensator for en- hancement of low voltage ride through (LVRT) capabil- ity of wind farms and active power flow management along with grid current harmonic mitigation. The pro- posed TCLC controller designed on generalized reactive power theory, enables the compensator to deliver reac- tive power in case of severe voltage dip during fault condition. It also controls the flow of active power under normal circumstance. In the presence of har- monics in the utility grid, the TCLC compensator can act as harmonic compensator. The compensator ele- ments have been designed depending on the range of injected reactive power and harmonic current rejection analysis. Simulation results and comparative assessme nt establish the improved performance of the compen- sator over other state-of-the-art techniques for LVRT improvement. Results obtained from the real-time dig- ital simulator (RTDS) prove the efficacy and reliability of the TCLC compensator.
A Solution map for extended economic load dispatch problem by secretary bird algorithm
(Vysoká škola báňská - Technická univerzita Ostrava, 2026) Pham, Ly Huu; Dinh, Bach Hoang; Phan, Tai Thanh; Dang, Tuu Kim; Luong, Bao Thien; Giang, Trung Thanh Nguyen
The paper introduces three applied methods - Secretary Bird Optimization Algorithm (SBOA), Par- ticle Swarm Optimization (PSO), and Tunicate Swarm Algorithm (TSA) - to address economic load dispatch problem (ELD) and the extended ELD problem with renewable energy resources (RES_ELD). These meth- ods were rigorously evaluated using various test systems with complex restrictions and objective functions. The test cases were ranged from simple to complex, with the most challenging involving load demands ranging from the minimum to the maximum load demand based on the total power of all units. The study’s results indi- cated that SBOA consistently outperformed PSO and TSA across all test systems, offering the best cost so- lutions in a shorter time. Also, SBOA demonstrates comparable or superior results as well as improved searchability compared to previous methods. Further- more, comparing these results highlighted SBOA’s ef- fectiveness in solving these problems and its potential for addressing engineering problems beyond ELD. Fi- nally, the study aimed to provide valuable insights for operators by suggesting solution map that operators can use it to make quick decisions to ensure safe and effi- cient system operation when generating capacity from power plants quickly meets load demand

Browse

Recent Submissions