ITEGAM-JETIA

An Effective GTO Algorithm-Based Cost-Benefit Analysis of DISCOs by Optimal Allocation of DG and DSTATCOM in a Radial Distribution Network

RAM PRASAD KANNEMADUGU, V. Adhimoorthy, A. Lakshmi Devi — Wed, 29 Jan 2025 12:05:48 -0300

Electric power Distribution Companies (DISCOs) is playing a major role for delivering active power from distribution substations to customers with lover cost, high reliability and voltage stability. In the DICOs, the transmission lines are redial in nature; all buses are containing the load and no generating buses. Therefore, voltage at each bus is minimized, loss of the network and voltage deviation is increased, and cost and benefit of the DISCOs and consumers are minimized. This paper maximizes the cost–benefit, and voltage stability of DISCOs is improved by optimally considering the Distributed Generation (DGs) and DSTATCOM. Here, applied a novel comprehensive Group Teaching Optimization (GTO) algorithm for planning DG units and DSTATCOM which considers both the Distribution Company’s and the DG Owner’s (DGO) profits simultaneously. The proposed GTO is applied to a 33-node test system and simulations are carried out using MATLAB platform and results show the applicability of the GTO in the DISCOs.

A A Fingerprint-Based Attendance System for Improved Efficiency

Olayiwola Charles Adesoba, Israel Mojolaoluwa Joseph — Wed, 29 Jan 2025 12:10:28 -0300

This paper presents the design and implementation of a fingerprint-based attendance system to address challenges in lecture attendance monitoring in developing countries. Leveraging a handheld fingerprint sensor, the proposed system streamlines attendance recording, eliminating manual collection inefficiencies and enhancing record reliability. The system enables lecturers to create and manage attendance sessions effortlessly, while students register and verify attendance conveniently. Key features include automated attendance tracking, reduced administrative burden, and improved accuracy. The system's successful deployment demonstrates its potential to improve operational efficiency and educational outcomes in resource-constrained environments. Results show significant reductions in attendance management time (by 75%) and errors (by 90%), alongside increased student accountability. User feedback indicates high satisfaction rates (95%). The system's effectiveness, usability, and scalability are discussed, highlighting its potential for widespread adoption. This research contributes to the development of efficient and reliable attendance monitoring solutions, providing valuable insights for educational institutions seeking to adopt biometric technology.

Comparative evaluation between Java application using JNI and native C/C++ application running on an Android platform.

Wed, 29 Jan 2025 12:16:08 -0300

Android is a popular operating system based on the Linux kernel and has a Java-based framework. As it is built on Linux, it supports the development of applications written in C/C++, known as native applications. The Native Development Kit (NDK), along with the Java Native Interface (JNI), provides a solution for communication between Java applications and native C/C++ applications, resulting in a significant performance boost. This article evaluated the performance difference between Java applications using JNI with the NDK and native C/C++ applications, focusing on algorithms widely used in various areas such as automation, networking, telecom, cybersecurity, etc. We conducted sequence of executions initiated either through a graphical interface or via the Android Debug Bridge (ADB) command line, with timing performed by external hardware with its own firmware for this evaluation. Based on the results, we observed that in all test cases, the native application performs faster, except when there are variations related to process scheduling, which may rarely lead to a reversal of this pattern.

Classification of Prominent Cacao Pod Diseases Using Multi-Feature Visual Analysis and k-Nearest Neighbors Algorithm

Earl Clarence San Diego, Seph Gerald Rodrin, Edwin Arboleda — Wed, 29 Jan 2025 12:20:11 -0300

Cacao has been one of the most promising crops produced in the Philippines due to its increasing demand in various local and international markets. Although cacao production aspired to be heightened to cope with the global trend, several difficulties were still needed to be addressed in crop propagation, mainly due to disruptive diseases and pests. In response to this problem, the study devised an algorithm based on k-Nearest Neighbors that can detect whenever a cacao pod was infected with the three most prominent diseases: black pod rot, Monilia, and pod borer infestations. The machine training model was preceded with visual feature extraction of color and texture parameters representing the cacao pod samples. It was found that the fine k-Nearest Neighbors algorithm achieved the highest validation and testing accuracies of 93.44% and 96.67%, respectively. The study's outcome suggested the continuous practicality of fusing visual feature extraction processes with supervised machine learning to generate models that can be applied to improve agricultural methods.

Development of malaria diagnosis with convolutional neural network architectures: a CNN-based software for accurate cell image analysis.

Emrah ASLAN — Wed, 29 Jan 2025 00:00:00 -0300

This study emphasizes that early diagnosis and treatment of malaria is critical in reducing health problems and mortality from the disease, especially in developing countries where the disease is prevalent. Malaria is a potentially fatal disease transmitted to humans by mosquitoes infected by a blood parasite called Plasmodium. The traditional method of diagnosis relies on experts examining red blood cells under a microscope and is inefficient as it is dependent on expert knowledge and experience. Nowadays, machine learning methods that provide high accuracy are increasingly used in disease detection. In this paper, a Convolutional Neural Network (CNN) architecture is proposed to distinguish between parasitized and non-parasitized cells. In addition, the performance of the proposed CNN architecture is compared to pre-trained CNN models such as VGG-19 and EfficientNetB3. The studies were carried out using the Malaria Dataset supplied by the National Institute of Health (NIH), and our proposed architecture was shown to function with 99.12% accuracy. The results of the study reveal that it is effective in improving the accuracy of cell images containing Plasmodium. In addition, a software that predicts whether cell images are noisy or not has been developed.

A A three-phase induction motor dynamic framework regulated by predictive and intelligent optimizations.

Shaswat Chirantan, Bibhuti Bhusan Pati — Wed, 29 Jan 2025 13:57:36 -0300

The role of Model Predictive Control (MPC) as a fundamental optimization tool in modern control systems is increasingly emphasized. In this context, the paper presents Predictive Current Control (PCC) strategies for a three-phase inverter-fed induction motor drive (IM), focusing on two core approaches: the Finite Control Set (FCS) and the Integral Finite Control Set (IFCS). The FCS-MPC algorithm is based on the evaluation of a cost function, selecting a control signal from a finite set that satisfies the minimum value of the cost function. This cost function is calculated based on the squared error between the reference current and the measured stator current. Conversely, the I-FCS-MPC uses a cascade feedback structure with an appropriately adjusted controller gain to determine the optimal set of control variables. Using a minimization principle, these methods manage the switching states for reversal, causing the inverter to generate appropriate voltage signals for the induction motor. This article compares IM electromagnetic torque and load currents under each control technique to determine the most flexible and robust prediction strategy. All these methods were studied in the MATLAB/Simulink environment. In addition, the paper uses Gravitational Search Algorithm (GSA) and Genetic Algorithm (GA) as benchmarks and shows that the results of FCS and I-FCS methods have superior performance.

The influence of the geometric features of processed surfaces on contact interaction and process performance during machining with elastic polymer-abrasive wheels.

Dmitriy Podashev — Wed, 29 Jan 2025 14:01:56 -0300

Автоматизация операций доводки и зачистки заусенцев остается весьма актуальной задачей для современного машиностроения. В данной статье рассматривается исследование влияния специфики контактного взаимодействия различных полимерно-абразивных кругов на производительность процесса обработки с целью определения зависимости между геометрической формой обрабатываемой поверхности и производительностью процесса обработки. Для теоретических расчетов и экспериментальных исследований использовались эластичные полимерно-абразивные круги фирмы 3М моделей FS-WL, DB-WL и CF-FB. Экспериментальные исследования проводились с использованием современного робототехнического комплекса на базе промышленного робота KUKA KR 210 R2700 EXTRA. Рассмотрены схемы взаимодействия кругов с различными поверхностями и для каждой из них определены формулы, позволяющие рассчитать среднюю деформацию и длину зоны контакта. Доказано влияние средней деформации и длины зоны контакта на эффективность процесса обработки. Полученные результаты необходимо учитывать при оптимизации рассматриваемых операций, а также при проектировании технологических процессов доводки деталей эластичным полимерно-абразивным инструментом.

Enhanced Performance of Microstrip Antenna Arrays through Concave Modifications and Cut-Corner Techniques

Wed, 29 Jan 2025 00:00:00 -0300

This paper presents the design and analysis of a high-performance 4×1 linear microstrip-fed antenna array optimized for wireless communication systems operating at 2.45 GHz. A novel concave-shaped modification is introduced on both the horizontal and vertical edges of the rectangular patch elements, significantly enhancing key performance metrics such as gain, impedance matching, and radiation efficiency. In addition, cut-corner techniques are applied to each patch element to minimize return loss and improve bandwidth, effectively addressing common limitations of traditional rectangular patch antennas, such as low gain and narrow bandwidth. Through rigorous simulations and physical prototyping, the proposed antenna array demonstrates a peak gain of 18 dB and a return loss of -33.82 dB at the target frequency. This makes it highly suitable for high-performance wireless applications, including WLAN, mobile communications, and smart transportation systems. The design not only improves antenna efficiency but is also cost-effective and simple to fabricate, making it ideal for mass production in modern communication systems.

Solving non-binary constraint satisfaction problems using GHD and restart.

Fatima AIT HATRIT, Kamal AMROUN, Professor — Wed, 29 Jan 2025 14:26:46 -0300

The non-binary instances of the Constraint Satisfaction Problem (CSP) could be efficiently solved if their constraint hypergraphs have small generalized hypertree widths. Several algorithms based on Generalized Hypertree Decomposition (GHD) have been proposed in the literature to solve instances of CSPs. One of these algorithms, called Forward Checking based on Generalized Hypertree Decomposition (FC-GHD+NG+DR), combines the advantages of an enumerative search algorithm with those of Generalized Hypertree Decomposition. However, like all structural decomposition methods, FC-GHD+NG+DR depends on the order in which the clusters are processed. In this paper, we propose a new version of the FC-GHD+NG+DR algorithm with a restart technique that allows changing the order of the nodes of GHD to improve performance. The experiments carried out are very promising, particularly on the satisfiable instances where we achieved better results using the restart method in 52.63% of the modified Renault satisfiable benchmarks and an average time resolution of for the normalized Pret and normalized Dubois benchmarks.

IoT-based location alert and controlling system for animal belts via mobile devices.

Vijay VIJAY Mane, Harshal Dirge — Wed, 29 Jan 2025 14:35:25 -0300

Traditional pet containment methods often lack efficiency and ease of use, posing significant challenges to maintaining pets within designated boundaries. This study presents the Animal Belt, an innovative geofencing-enabled pet management system designed to monitor and control pets' movements. The system utilizes GPS technology to establish virtual boundaries, triggering vibratory feedback and pre-recorded voice commands when the pet breaches the defined geofence. Additionally, owners receive real-time alerts via a mobile application, including a Google Maps link for precise pet location tracking. Experimental evaluations validated the system’s performance within a 300-meter geofence radius, demonstrating consistent activation of feedback mechanisms upon boundary violations. The results underscore the system’s ability to enforce geofence limits effectively, leveraging feedback mechanisms to prompt pets' return. Key features include customizable operational settings and a user-friendly interface, offering a modern alternative to traditional leashing methods. The proposed system enhances pet safety, minimizes owner intervention, and provides a reliable solution for outdoor pet management.

Fault diagnosis and fault-tolerant control strategy for interleaved boost DC/DC converter dedicated to PEM fuel cell applications.

Wed, 29 Jan 2025 14:49:03 -0300

This paper proposes an improved fault diagnosis and fault tolerant control (FTC) strategy for interleaved boost DC/DC converter that is suitable for fuel-cell applications.

This paper investigates a two-phase interleaved boost DC-DC converter. This design offers several advantages, including:

Low ripple current, by splitting the load current between two phases, the ripple current at the input and output is significantly reduced compared to a single-phase converter.

Reduced semiconductor stress, Each phase handles only a fraction (1/N) of the total current, which reduces stress on individual components and promotes higher reliability and operating margins.

Furthermore, the paper proposes and evaluates an H-infinity controller for the converter. This advanced control strategy ensures robust performance despite variations in reference voltage and load conditions.

The power converter suffers from failure switching due to various factors. To address these drawbacks and achieve both accurate reference tracking with desired dynamic response and rapid fault detection, an algorithm based on current-slopes are proposed.

Minimizing current ripples is crucial to ensure the longevity of PEMFCs, so the interleaved boost converter structure is dedicated to the PEMFCs in order to reduces the ripple of the generated current.

The overall system has been simulated using MATLAB/Simulink software under different conditions such as reference voltage variation, load variation, and Short Circuit default ; the obtained results in different phase demonstrate the higher performance, of the proposed systems in terms of dynamic performance, fast fault detection and fault tolerant action to restore the health stat.

Transformer-Based Optimization for Text-to-Gloss in Low-Resource Neural Machine Translation

Younes Ouargani, Noussaim El Khattabi — Wed, 29 Jan 2025 00:00:00 -0300

Sign Language is the primary means of communication for the Deaf and Hard of Hearing community. These gesture-based languages combine hand signs with face and body gestures for effective communication. However, despite the recent advancements in Signal Processing and Neural Machine Translation, more studies overlook speech-to-sign language translation in favor of sign language recognition and sign language to text translation. This study addresses this critical research gap by presenting a novel transformer-based Neural Machine Translation model specifically tailored for real-time text-to-GLOSS translation. First, we conduct trials to determine the best optimizer for our task. The trials involve optimizing a minimal model, and our complex model with different optimizers; The findings from these trials show that both Adaptive Gradient (AdaGrad) and Adaptive Momentum (Adam) offer significantly better performance than Stochastic Gradient Descent (SGD) and Adaptive Delta (AdaDelta) in the minimal model scenario, however, Adam offers significantly better performance in the complex model optimization task. To optimize our transformer-based model and obtain the optimal hyper-parameter set, we propose a consecutive hyper-parameter exploration technique. With a 55.18 Recall-Oriented Understudy for Gisting Evaluation (ROUGE) score, and a 63.6 BiLingual Evaluation Understudy 1 (BLEU1) score, our proposed model not only outperforms state-of-the-art models on the Phoenix14T dataset but also outperforms some of the best alternative architectures, specifically Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), and Gated Recurrent Unit (GRU). Additionally, we benchmark our model with real-time inference tests on both CPU and GPU, providing insights into its practical efficiency and deployment feasibility.

The Advances in Neuromorphic Computing and Brain-Inspired Systems (ANCBIS).

DanielRaj K, Ponseka G, Bharath Sanjai Lordwin D J3 — Wed, 29 Jan 2025 15:00:54 -0300

Neuromorphic computing, inspired by the structure and functions of the human brain, is transforming the development of energy-efficient, adaptive, and highly parallel processing systems. This field seeks to bridge the gap between traditional computing architectures and biological neural networks by replicating brain-like functionalities. This paper examines recent advancements in neuromorphic computing, with an emphasis on innovative hardware and algorithms that boost computational power while reducing energy consumption. Key technologies such as memristive devices, spiking neural networks, and brain-inspired learning algorithms show promise in applications like pattern recognition, sensory processing, and autonomous decision-making. This study also addresses challenges related to scalability, robustness, and integration with existing systems, emphasizing the importance of cross-disciplinary collaboration to overcome these limitations. By exploring applications in robotics, medical diagnostics, and environmental monitoring, this research highlights how brain-inspired systems could drive the next generation of artificial intelligence and sustainable computing, meeting the growing need for energy-efficient, intelligent systems.

From Backtracking To Deep Learning: A Survey On Methods For Solving Constraint Satisfaction Problems

Fatima AIT HATRIT — Wed, 29 Jan 2025 15:05:06 -0300

Constraint Satisfaction Problems (CSP) are a fundamental mechanism in artificial intelligence, but finding a solution is an NP-complete problem, requiring the exploration of a vast number of combinations to satisfy all constraints. To address this, extensive research has been conducted, leading to the development of effective techniques and algorithms for different types of CSPs, ranging from exhaustive search methods, which explore the entire search space, to modern techniques that use deep learning to learn how to solve CSPs. This paper represents a descriptive and synthetic overview of various CSPs solving methods, organized by approach: systematic search methods, inference and filtering methods, structural decomposition methods, local search-based methods, and deep learning-based methods. By offering this structured classification, it presents a clear view of resolution strategies, from the oldest to the most recent, highlighting current trends and future challenges, there by facilitating the understanding and application of available approaches in the field.

Enhanced Brain Tumor MRI Classification Using Stationary Wavelet Transform, ResNet50V2, and LSTM Networks

Oussama Abda, Hilal NAIMI — Wed, 29 Jan 2025 00:00:00 -0300

Brain tumors constitute a significant health issue in the world today because of their aggressive behavior and short survival rates. Early and accurate detection of brain tumors is necessary for effective treatment and improved patient outcomes. The principal diagnostic technology that shows highly detailed visualization of brain structures is Magnetic Resonance Imaging (MRI); however, the interpretation of these images can be time-consuming and require expertise and highly specialized manpower. This study presents a new approach for brain tumor classification, which combines advanced preprocessing, feature extraction, and classification techniques. The preprocessing includes Stationary Wavelet Transform (SWT) intended to enhance tumor-relevant features and resizing to standard MRI image dimensions; feature extraction includes. After that a Long Short Term Memory network receives the features. that will model the dependencies in the feature space and classifies into four categories: Glioma, Meningioma, Pituitary tumors, and No Tumor. Experiments showed that this proposed method can be effective in producing a high classification accuracy rate along with time quality processing. This work brought forward the prospects of developing an automated, accurate, and reliable brain tumor classification system from SWT, ResNet50V2, and LSTM, whereas otherwise, it catered for needs in the enhancement of diagnostic tools in medical imaging. The method was analyzed using the Kaggle dataset and scored an amazing accuracy of 98.7%, which proved the effectiveness of the method in improving brain tumor classification.

Enhancing Medical Education: Building a Comprehensive E-Learning Platform with CodeIgniter 4

Meftah Zouai, Ahmed ALOUI, Houcine BELOUAAR, Ilyes Naidji, Okba KAZAR — Fri, 21 Feb 2025 12:00:00 -0300

The emergence of the COVID-19 pandemic has presented unprecedented difficulties for medical education, forcing institutions worldwide to adjust quickly to ensure that learning continues despite the implementation of restrictive measures and social distancing procedures. This article explores creating and implementing a cutting-edge e-learning platform designed exclusively for medical education. Utilising the CodeIgniter 4 PHP framework for backend development and Bootstrap for frontend design, the platform provides a wide range of interactive quizzes, including Multiple Choice Questions (QCM), Single Choice Questions (QCU), and clinical cases (Cas Clinique’s). The platform's adaptable design enables medical students to easily access and engage in remote learning across different platforms, allowing them to continue their education without interruption. The main characteristics consist of instruments for analysing performance, allowing students to track their progress and personalise their study sessions, thus improving the effectiveness and adaptability of medical education. This article highlights the significant impact of e-learning in addressing the educational challenges caused by the COVID-19 pandemic and provides insights into the future of medical education. It achieves this by thoroughly examining the platform's architecture, features, and pedagogical implications.

Predicting Remaining Useful Life of Lithium-Ion Batteries for Electric Vehicles Using Machine Learning Regression Models

Sravanthi C L, Dr.J N Chandra sekhar — Fri, 21 Feb 2025 00:00:00 -0300

Accurate prediction of a lithium-ion battery's remaining useful life (RUL) is essential for effectively managing and maintaining electric vehicles (EVs). By anticipating battery health and potential failures, we can optimize performance, enhance safety, and prevent costly breakdowns. Based on a supervised machine-learning regression approach, this work presents four different regression models like Gradient Boosting Regressor, K-Nearest Neighbor Regressor, Bagging Regressor, and Extra Tree Regressor models to forecast the li-ion battery life for electric vehicles. Using actual battery data from Hawaii National Energy Institute (HNEI), four algorithms were used to forecast remaining useful life (RUL) of batteries. These algorithms were implemented using Python in Google Co-laboratory. The accuracy of each model, Performance error indices including Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), R-squared, and computational time were calculated. Findings show that Bagging Regressor model outperforms the other three models in terms of RUL prediction. The Bagging Regressor model demonstrated its superiority with better

Parametric Analysis of UFMC with 5G NR Polar and Convolutional Codes in a Massive MIMO System

Smita Prajapati, Divya Jain, Neha kapil — Fri, 21 Feb 2025 14:52:48 -0300

The Fifth Generation (5G) wireless network's radio access strategies must meet dynamic and adaptable service requirements. The major demands in the current era of pervasive wireless networks are high throughput, reliability, and secure connectivity. 5G New Radio (NR) air interface is a major transition to new modulation and channel coding techniques to reduce redundancy, latency, and complexity. Convolutional codes were used in 4G and polar codes in 5G to code channels for control information in the uplink and downlink. This research aims to investigate the 4G channel codes and provide analytical results for comparing them to the 5G polar codes in Ultra-Reliable Low-Latency Communication (URLLC) applications with short block-length transmissions. The research implements Universal Filtered Multi-Carrier (UFMC) modulation, a suitable technique for short burst transmissions. Channel coding is applied to enhance reliability, considering Polar codes as major 5G candidates for short packet transmission. The comprehensive system is simulated in a massive Multiple Input Multiple Output (MIMO) scenario. The impact of antenna array size in MIMO and UFMC parameters and sub-band size are investigated. The major contribution of the work is that the Bit Error Rate (BER) performance of Polar codes is enhanced with an SNR gain of ~7dB with a 64x16 MIMO UFMC system compared to convolutional codes. Moreover, the concatenated polar and convolutional codes are used, which results in an additional SNR boost of about 3dB. This research reveals that mission-critical applications in 5G can benefit from the flexibility and improved error rate performance offered by the combination of UFMC, Polar codes, and massive MIMO.

A Logistics 5.0 maturity model: a human-centric and sustainable approach for the supply chain of the future

Nazare Toyoda Machado, Carlos Manuel Taboada Rodriguez — Fri, 21 Feb 2025 15:27:17 -0300

A Logística 5.0 representa um avanço transformador ao integrar tecnologias avançadas, como inteligência artificial, aprendizado de máquina e blockchain, com uma abordagem centrada no ser humano e focada na sustentabilidade. Ao contrário da Logística 4.0, que priorizou a automação e a digitalização, a nova abordagem enfatiza a colaboração entre humanos e máquinas para criar cadeias de suprimentos mais eficientes e resilientes. Os modelos de modernidade específicos para esta fase emergente são cruciais para avaliar a prontidão tecnológica, a capacidade de integração homem-máquina e o compromisso com práticas sustentáveis. A aplicação de tecnologias colaborativas como veículos autônomos, algoritmos preditivos e robôs otimiza processos, reduz erros e minimiza impactos ambientais, alinhando-se às metas globais de sustentabilidade. Além disso, práticas logísticas verdes, como o uso de energia renovável e a economia circular, são essenciais para reduzir a pegada de carbono das empresas. No entanto, a transição para a Logística 5.0 enfrenta desafios significativos, incluindo a necessidade de investir em infraestrutura, treinar funcionários e superar barreiras culturais. No entanto, as empresas que adotam essa abordagem não apenas aumentam sua competitividade, mas também se voltam para uma economia mais sustentável e resiliente, posicionando-se à frente em um mercado global vibrante e voltado para a inovação.

A Measurement Model of Logistics 5.0 Maturity: An Integrative Review and Framework Proposal Based on Literature

Nazare Toyoda Machado, Carlos Manoel Taboada Rodriguez — Fri, 21 Feb 2025 15:34:36 -0300

This article explores Logistics 5.0 as an evolution of Industry 4.0, emphasizing the integration of emerging technologies such as artificial intelligence, IoT, and big data in logistics management. It proposes a specific maturity measurement model for Logistics 5.0, structured into five levels: initial, repeatable, defined, managed, and optimized, which evaluate technological readiness, process management, analytical capacity, change management, and sustainability. The analysis highlights gaps in traditional models, proposing dimensions adapted to the demands of digital transformation. The model emphasizes the harmonization between technology, people, and processes, pointing toward more efficient, adaptable, and sustainable logistics. It concludes that the practical application of the model can help companies enhance their competitiveness and sustainability in a dynamic global market.

Parametric study of the thermal behavior of cold metal transfer welding with titanium

Mohamed Walid Azizi, Djoubeir Deddah, Ibtissem Gasmi — Fri, 21 Feb 2025 15:46:38 -0300

This work focused on enhancing the efficiency of thermal behaviour in the application of Cold Metal Transfer (CMT), especially for welding tough metals like titanium, which has the potential to impact the field of welding technology significantly. The investigation of the thermal behavior of CMT welding, carried out by means of parametric analysis, was a crucial step in this direction. This study, which was carried out with the assistance of numerical simulations with COMSOL Multiphysics, particularly emphasized critical factors such as plate thickness and welding power. The significance of this study in advancing our understanding of additive manufacturing in welding is highlighted by the results of the study. These results, which illustrate the effects of the specified influencing parameters through temperature distribution at various time intervals, 2D and 3D graphs depicting temperature evolution along the welding path, and the 2D temperature profile at (t=5s) across different plate thicknesses, have the potential to revolutionize the field of welding technology and bring about exciting new possibilities.

Inter-Cluster Distance-Based SMOTE Modification for Enhanced Diabetes Classification

Fri, 21 Feb 2025 16:05:14 -0300

Diabetes is a significant global health challenge, with early diagnosis playing an important role in preventing serious complications. However, medical datasets often exhibit class imbalance, where the number of non-diabetes cases is much larger than diabetes cases. This imbalance causes machine learning models to be biased towards the majority class, thus degrading prediction performance on the minority class. The problem with the commonly used oversampling method SMOTE (Synthetic Minority Oversampling Technique) is that the selection of new synthetic data formation points is done randomly, which often results in less representative synthetic data and reduces model performance. This research proposes a modification of SMOTE based on inter-cluster distance to overcome this problem. This approach uses the distance between cluster centroids in minority classes to form new synthetic data that is more representative. The research methodology involves data preprocessing, including missing value imputation, normalization, and data balancing using SMOTE modification, followed by classification using Random Forest algorithm. Evaluation was conducted using accuracy, precision, recall, and F1-score metrics. The results showed that the proposed approach achieved very high evaluation values, with accuracy, precision, recall, and F1-score of 99.7% each, far surpassing previous studies that used standard oversampling methods. This study proves that the inter-cluster distance-based SMOTE modification is effective in overcoming class imbalance and producing more representative synthetic data.

Artificial Neural Network-Based Deadbeat Predictive Current Control with Dead-Time Compensation for PMSMs

Fri, 21 Feb 2025 16:15:23 -0300

In the velocity control of Permanent Magnet Synchronous Motors (PMSMs), Deadbeat Predictive Current Controllers (DPCCs) are renowned for their excellent dynamic performance and constant switching frequency. However, achieving precise velocity regulation remains challenging due to the nonlinearities introduced by two-level voltage source inverter (2L-VSI). Specifically, the dead time inherent in 2L-VSI results in voltage distortion, which generates parasitic harmonics in the system. These harmonics degrade control accuracy, cause a current ripple, and can lead to performance degradation or even system instability, compromising reliable operation. This article proposes an innovative solution: Artificial Neural Network-Based Deadbeat Predictive Current Control (ANN-DPCC) integrated with dead-time compensation to address these issues. This approach effectively suppresses the current ripple and significantly reduces total harmonic distortion (THD). Simulation results validate that ANN-DPCC with dead-time compensation outperforms traditional DPCC by improving response times, enhancing steady-state accuracy, and minimizing current distortions. This novel strategy significantly advances PMSM control, offering precise velocity regulation, improved reliability, and superior system performance for demanding applications

Optimizing Artificial Neural Networks with Particle Swarm Optimization for Accurate Prediction of Insulator Flashover Voltage Under Dry and Rainy Conditions

Abdelhalim Mahdjoubi, lazreg taibaoui, Boubakeur Zegnini — Fri, 21 Feb 2025 17:03:13 -0300

Outdoor insulators are highly susceptible to environmental factors, such as moisture, rain, and contaminants, which significantly degrade their efficiency and durability. These factors contribute to surface flashovers, leading to insulation failures in outdoor power systems. This study presents a novel application of advanced machine learning techniques to predict the flashover performance of glass insulators under diverse environmental conditions, focusing on dry and rainy scenarios. The research emphasizes the critical role of raindrops in reducing flashover voltage. A hybrid model combining Artificial Neural Networks (ANN) with Particle Swarm Optimization (PSO) is developed to address these challenges. The PSO algorithm optimizes the ANN's hyperparameters, enabling the model to establish a nonlinear relationship between key insulator characteristics, including standard and anti-pollution profiles and their critical flashover voltage. Rigorous testing demonstrated exceptional accuracy, with a mean absolute percentage error (MAPE) of 0.2458 and a near-perfect coefficient of determination (R²) of 0.999. These findings highlight the robustness and reliability of the proposed hybrid model in predicting flashover voltage under varying environmental conditions. This work provides a powerful tool for enhancing the design, maintenance, and operational reliability of outdoor insulators, particularly in regions prone to high levels of pollution and moisture, contributing significantly to the advancement of sustainable power transmission systems.