Master Thesis
Permanent URI for this collection
Browse
Recent Submissions
Item Open Access Control for the Fuel and Oxygen Flows of the Solid Oxide Fuel Cell by two Types of Fuzzy Adapve PID(University of Msila, 2025-06-30) BABAH MostafaSI ABDALLAH Marouane; SI ABDALLAH Marouane; ENCA/OUAGUENI Fayssal; CO.ENCA/HERIZI AbdelghafourThis study aims to enhance performance of Solid Oxide Fuel Cells (SOFCs) by precisely controlling fuel and oxygen flow using two types of fuzzy adaptive PID controllers: Type- 1 and Type-2. dynamic model of SOFC was developed and traditional PID controller was combined with both fuzzy logic types. Type-2 fuzzy controller showed superior capability in handling uncertainty and noise compared to Type-1, albeit at cost of higher computational complexity. Simulation results revealed that fuzzy adaptive control especially using Type-2 logic greatly improves system response time, eliminates overshoot and enhances dynamic stability. Optimization techniques like Anti-Windup and differential forward algorithms were also applied to improve robustness. These findings contribute to advancing reliable and efficient SOFC systems for real-world applications.Item Open Access Advanced Machine Learning for Heart Disease Prediction A Comparative Study of Ensemble and Traditional(University of Msila, 2025-06-30) BELAADA Wahid Akram; AFAN MohamedHeart disease, a leading global cause of mortality, demands innovative early detection strategies. This study evaluates machine learning models (DNN, KNN, SVM, XGBoost, Random Forest) for predicting cardiovascular disease using 1,000 patient records with 16 clinical features. After rigorous preprocessing and validation, ensemble methods like XGBoost (100% accuracy) and Random Forest (99%) outperformed traditional models, highlighting their clinical potential. Challenges such as overfitting and interpretability were addressed, emphasizing the need for diverse datasets and explainable AI (XAI). Future integration with wearable technologies and interdisciplinary collaboration could enable proactive, personalized care, transforming cardiovascular health outcomes globally.Item Open Access Analysis of Smallest-Of Cell Averaging CFAR Detection in Homogeneous Gamma-Distributed Background(University of M'sila, 2025-06-30) Mohamed DJABALLAH; ENCA/ Mohamed SAHEDThis study investigates the problem of adaptive radar target detection in both homogeneous and non-homogeneous Gamma-distributed clutter environments. The objective is to maintain a Constant False Alarm Rate (CFAR) during the detection process. It is assumed that the radar system employs a square-law detector preceding the CFAR processing stage. Initially, the fundamental principles of radar target detection in noise are presented, along with an overview of CFAR detection techniques. The study then focuses on a detailed examination of several Mean-Level CFAR detectors operating in Gamma-distributed clutter, specifically the Cell Averaging (CA-CFAR), Greatest Of (GO-CFAR), and Smallest Of (SO-CFAR) algorithms. A comprehensive theoretical analysis is conducted for each detector. Closed-form expressions for the probability of false alarm (Pfa) are derived. However, the calculation of the probability of detection (Pd), based on the exact statistical characterization of the cell under test (CUT), involves complex integrals that are computationally intensive. To address this issue, approximate expressions for Pd are proposed. These approximations offer computational efficiency and are suitable for real-time implementation. The theoretical findings are validated through numerical evaluation, including both integral-based computations and Monte Carlo simulations, under various clutter scenarios. Furthermore, a performance comparison between the studied detectors and the optimal detector is performed, assuming a homogeneous clutter environment. The results demonstrate that the SO-CFAR detector exhibits superior performance in homogeneous clutter, while the GO-CFAR detector proves to be more effective in nonhomogeneous environments.Item Open Access Étude des méthodes de protection des services dans les réseaux mobiles : application au réseau Mobilis dans la région de M'Sila.(UniversitY of M'sila, 2025-06-30) HALLAB Bilal; BELDI Khalil; ENCA/ GARAH MessaoudCe mémoire se focalisera sur l'étude des diverses technologies de transmission disponibles dans le cadre du développement des réseaux mobiles, tout en proposant une description pratique de l'équipement HUAWEI (micro-wave et fibre optique). Cette démarche est perçue comme une étape incontournable pour acquérir une expérience sur ce genre d'équipement. Au cours de notre stage, nous avons mené une action concrète concernant le réseau de transmission FH avec l'opérateur national de téléphonie mobile ATM MOBILIS, dans le but d'améliorer la qualité du service et d'éviter les interruptions du trafic par des appels interrompus. Suite à notre intervention, le complément E1, en raison d'une saturation dans le premier E1 (problème relevé par un CR), la capacité du service voix a été élargie (32 Time slots) et appliqué la protection SNCP, ce qui a résolu le problème du service vocal et a empêché les perturbations du trafic.Item Open Access Enhancing Brain Tumor Classification Accuracy through Multi-Modal MRI Analysis and Advanced Ensemble Deep Learning(UNIVERSITY OF M’SILA, 2025-06-30) BERRA Ayyoub; ENCA/ ATTALLAH BilalIn an attempt to apply state-of-the-art deep learning techniques for brain tumor classification, which is a medical imaging domain, accurate multi-class diagnosis can dramatically affect treatment choices and patient outcomes, two modern CNN architectures EfficientNet-B3 and ConvNeXt-Tiny were evaluated for the classification of 44 different brain tumor types through Magnetic Resonance Imaging (MRI) data. The dataset, sourced from Kaggle. To redress class imbalance and enhance the generalization aspect, several data augmentation methods . Transfer learning was implemented by employing pretrained weights, and both models were fine-tuned with customized classifier heads that have Batch Normalization, Dropout and Dense layers for training. The same hyperparameters were used to train all models, and the evaluation was done with Accuracy, F1-Score, Recall, and Precision. The results demonstrate that EfficientNet-B3 reached an accuracy of 96.85%, F1-Score 96.74%, Recall 96.85%, and Precision 96.94%, while ConvNeXt-Tiny got accuracy of 95.88% ,F1-Score 95.86%, Recall 95.88%, and Precision 96.16%. A weighted ensemble approach was also used to combine the two models. This ensemble model attained the highest classification accuracy of 97.07%, which shows that weighted averaging is a good way to improve performance. Such outcomes demonstrate the strong points of present deep learning architectures and ensemble methods in dealing with fine-grained multi-class brain tumor classification. The findings confirm that hybrid approaches have the capability to surpass accuracy in addressing difficult diagnostic tasks from medical imaging data.Item Open Access Étude des structures plasmoniques MIM pour la réalisation des capteurs de température(University of M’sila, 2025-06-30) BOUZID Saadi; BOUZID Nadjib; ENCA/ HOCINI AbdesselamCe mémoire traite de la conception et de l’optimisation d’un capteur de température basé sur la résonance plasmonique de surface (SPR), utilisant une structure métal-isolant-métal (MIM). Les simulations ont été réalisées à l’aide de la méthode FDTD (Finite-Difference Time-Domain) via le logiciel RSoft/FullWAVE. L’objectif principal est d’exploiter les propriétés de confinement des modes plasmoniques pour améliorer la sensibilité du capteur. Différentes configurations géométriques ont été étudiées, notamment par l’introduction de cavités résonantes dans la structure. Les résultats obtenus ont montré une amélioration significative de la sensibilité, atteignant jusqu’à 1500 nm/RIU, ce qui confirme la pertinence de l’approche adoptée pour des applications de détection optique de haute précision.Item Open Access Solar Energy Prediction using PSO_SVR - Case study Adrar site -(University of M'sila, 2025-06-30) MAHDID Sabah; ENCA/ Mezaache HatemThe integration of renewable energy sources into the energy mix is crucial to the transition to a low-carbon economy. Among these sources, solar energy occupies an important place due to its availability and potential to provide clean, renewable energy. However, the intermittent and variable nature of solar power generation creates major challenges for the management and stability of power grids. This work proposes an innovative approach for estimating solar power over different time horizons, combining advanced optimization and machine learning techniques. Particle swarm optimization (PSO) is used to efficiently adjust the hyper-parameters of support vector regression (SVR), thereby improving the accuracy of the estimates. This method better captures variations in solar production as a function of meteorological and temporal conditions, offering a powerful and adaptable solution for energy forecasting applications. To evaluate our estimation system two evaluation methods are employed, a statistical evaluation based on performance indicators such as root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R²), as well as a graphical evaluation using scatter plots to compare actual data with predicted results, these evaluation methods allowing to compare predictions with actual values and to verify the effectiveness of the proposed hybrid approach.Item Open Access Commande vectorielle par Backstepping de la machine asynchrone pentaphasée(UNIVERSITY OF M’SILA, 2025-06-30) KHEZZARI DOUNYA; NEBBATI DOUNIA; ENCA/ FODIL MALIKALa commande vectorielle par Backstepping appliquée à la machine asynchrone pentaphasée représente une stratégie avancée permettant d’exploiter pleinement le potentiel des systèmes multi-phasés. Grâce à la commande vectorielle, les grandeurs électriques sont transformées dans un repère synchronisé, ce qui permet un découplage clair entre le couple et le flux, similaire à celui observé dans les machines à courant continu. Cette transformation simplifie la commande dynamique et améliore la précision du contrôle. De son côté, la méthode Backstepping, adaptée aux systèmes non linéaires, permet de concevoir des lois de commande basées sur la stabilité de Lyapunov, en construisant progressivement des contrôleurs à chaque étape du système. L’association des deux techniques permet d’obtenir une commande robuste, stable et performante face aux perturbations et incertitudes du système. Elle est particulièrement adaptée aux machines pentaphasées, qui présentent des caractéristiques dynamiques complexes et une forte tolérance aux défauts, notamment dans les applications industrielles et de traction électrique.Item Open Access FUZZY - SMC - PI Regulator for Speed Control of the Doubly Fed Induction Machine(University of M’sila, 2025-06-30) SAOUCHI Louey; BELABBES Fares; ENCA/ HERIZI AbdelghafourThis thesis proposes a robust nonlinear control strategy for the Doubly-Fed Induction Machine (DFIM), combining PI control, Sliding Mode Control (SMC), and type-1 fuzzy logic. The goal is to improve system performance under variable speeds and loads. Various control methods, including vector control and SMC, were evaluated through simulations. Results confirm the dynamic advantages of the hybrid fuzzy controller in enhancing speed, torque, and flux regulation.Item Open Access Performance of RIS-empowered communication networks over symmetric/asymmetric fading channels(University of M’sila, 2025-06-30) SAKOU Mounsif; DEFAF Abdelwahab; ENCA/ BENMAHMOUD SlimaneLa demande croissante en débits de données élevés, en services de communication à très haut débit et en couverture étendue représente un défi majeur pour la conception des futurs systèmes sans fil. Parmi les solutions prometteuses pour répondre à ces exigences, les surfaces intelligentes reconfigurables (Reconfigurable Intelligent Surfaces, RIS) se démarquent par leur potentiel. Ce mémoire examine les performances en termes de probabilité de panne (outage) et de capacité ergodique des systèmes de communication sans fil assistés par des RIS. À travers l’analyse des indicateurs OP (Outage Probability) et EC (Ergodic Capacity), nous évaluons l’impact de différents paramètres système sur les performances globales du réseau. Les résultats obtenus montrent que l’intégration des RIS permet une amélioration significative de ces deux indicateurs, mettant ainsi en évidence leur efficacité dans l’optimisation des communications sans fil. Des simulations approfondies démontrent également que les performances du système sont fortement influencées par certains paramètres, en particulier le nombre d’éléments constituant la surface RIS. Ces observations soulignent l’importance cruciale d’un bon paramétrage des RIS pour garantir une qualité de signal optimale et une fiabilité accrue du réseau, ouvrant ainsi la voie au développement des réseaux de communication sans fil de nouvelle génération.Item Open Access Integrated learning system for industrial automation and instrumentation Based on PID-Controller Using Tia Portal and Factory I/O(University of M’sila, 2025-06-30) RAGGAI Abdeloihab; ENCA/ GHELLAB Mohamed ZinelaabidineThe thesis titled "Integrated Learning System for Industrial Automation and Instrumentation: Based on PID-Controller Using TIA Portal and Factory I/O" aims to bridge the gap between theoretical knowledge and practical application in the field of industrial automation by designing and implementing a closed-loop control system to regulate the water level in a tank using a PID controller. The TIA Portal platform was used for PLC programming and HMI design, and it was integrated with a 3D simulation using Factory I/O, providing a realistic virtual environment. After analyzing the system's open-loop response and applying the Ziegler–Nichols method to determine tuning parameters, the PID controller was activated and tested under various setpoint changes and disturbances. Results showed that the PI controller provided smoother and more stable responses, making it the most suitable for this application. The project demonstrates the importance of combining theoretical knowledge with hands-on skills using modern industrial tools.Item Open Access Détection d'interaction humaine-objet à l'aide du deep learning(UNIVERSITY OF M’SILA, 2025-06-30) Benslimane Manel; Halitim Amira; ENCA/ LALAOUI LahouaouiLa détection d'interaction humain-objet, un domaine clé de la vision par ordinateur, en mettant l'accent sur les avancées apportées par les techniques de Deep learning. Il explore diverses architectures de réseaux de neurones, telles que R-CNN et YOLO, et analyse leur efficacité dans l'identification et le suivi d'objets dans des environnements complexes. En abordant les défis liés à la détection, comme les occlusions et les variations d'éclairage, le mémoire démontre l'importance d'une approche intégrée pour améliorer la précision et la fiabilité des systèmes de détection dans des applications pratiques.Item Open Access Diagnostic et commande tolérante aux défauts d’un système hydraulique à quatre réservoirs(University of Msila, 2025-06-30) BEN GANA NABIL; CHARIF AHMED; Rappo/OUBABAS HocineCe mémoire de fin d'étude porte sur le diagnostic et la commande tolérante aux défauts d'un système hydraulique à quatre réservoirs. Après une brève description du système à quatre réservoirs, son modèle mathématique non linéaire et linéarisé sont calculés. un régulateur PID est synthétisé pour améliorer les performances du système. Afin de détecter les défauts actionneurs, l'observateur à entrées inconnues est choisit comme générateur de résidus. Après détection et localisation des défauts, une loi de commande reconfigurable est proposée pour améliorer les performances dégradées du système en présence de défauts.les résultats sont illustrés par simulation.Item Open Access Améliorer l'accessibilité pour les malvoyants : Une solution basée sur l'IA(University of Msila, 2025-06-30) Bachiri Bochra; Alizouaoui Sabrina; Benattig Allaa; Meftah Zahra; ENCA/BRIK MouradL'apprentissage automatique suscite beaucoup d'attention ces derniers temps dans le domaine de la vision par ordinateur, car il évolue constamment. Les avancées de l'apprentissage automatique ont été utilisées dans de nombreux domaines, y compris dans le développement de solutions d’assistance pour les personnes malvoyantes. Parmi les défis auxquels ces personnes sont confrontées, on trouve la difficulté d'accéder aux informations visuelles dans leur environnement, ce que nous avons abordé dans ce mémoire. Dans ce travail, nous présentons un système intelligent permettant aux utilisateurs malvoyants de détecter et de lire les panneaux médicaux de manière vocale, renforçant ainsi leur autonomie lors de leurs déplacements. L'application utilise un modèle YOLO pour détecter et classifier en temps réel les panneaux à partir des images capturées par la caméra. En plus de la détection des panneaux, le système extrait le texte des images à l'aide de la technologie de reconnaissance optique de caractères (OCR), puis le convertit en parole via un module de synthèse vocale (TTS), afin que l'utilisateur puisse écouter instantanément l'information détectée. Ce travail constitue une avancée significative vers le développement de solutions intelligentes basées sur l'intelligence artificielle, visant à améliorer la qualité de vie des personnes malvoyantes, tout en ouvrant des perspectives prometteuses pour de futures évolutions dans ce domaine.Item Open Access Filter Bank Multicarrier Modulation Techniques for 5G Communication Systems(University of M’sila, 2025-06-30) Cheriat Mohamed; Achraf Ziam Saad Eddine; ENCA/ ZERDOUMI ZohraThis thesis investigates multicarrier modulation techniques for 5G telecommunications, comparing Orthogonal Frequency Division Multiplexing (OFDM) and Filter Bank Multicarrier (FBMC). It addresses 4G limitations, such as spectral efficiency and interference, and evaluates FBMC’s potential for 5G. Using MATLAB simulations, the study compares power spectral density (PSD), peak-to-average power ratio (PAPR), and bit error rate (BER). Results show FBMC’s superior spectral efficiency and lower PAPR, positioning it as a strong 5G candidate, though OFDM’s simplicity is advantageous in some cases. Future research on FBMC is recommended to optimize 5G performance.Item Open Access Candidate Waveforms in Future Wireless Communication Systems(University of M’sila, 2025-06-30) AOUINA Said; MELIANI Elmahdi; ENCA/ ZERDOUMI ZohraThe rapid evolution of wireless communication systems has increased the need for advanced waveform designs that can support higher data rates, lower latency, and better spectral efficiency. Among several candidates, Universal Filtered Multicarrier (UFMC) has emerged as a promising solution due to its ability to minimize out-of-band emissions and operate effectively in asynchronous and fragmented spectrum scenarios. In this thesis, we evaluate UFMC in comparison to the conventional Orthogonal Frequency Division Multiplexing (OFDM), which has been the dominant waveform in 4G and 5G systems. The study includes a detailed simulation-based performance analysis using MATLAB. We investigate key performance metrics such as Power Spectral Density (PSD), Peak-to-Average Power Ratio (PAPR), Out-of-Band Emissions (OOBE), Bit Error Rate (BER), and Symbol Error Rate (SER) under various modulation schemes (4-QAM to 256-QAM) and a Rician fading channel. The Simulation results demonstrate that UFMC achieves better spectral containment, more stable PAPR behavior with high-order modulations, and improved OOBE suppression outperforming OFDM in several scenarios relevant to 5G and future 6G requirements. This work highlights UFMC’s advantages and limitations and confirms its potential as a strong waveform candidate for next-generation wireless systems.Item Open Access Trimodal Generalized Gamma Distribution of Sea Echoes and CFAR Detection in CG-LNT Clutter with Multiple Order Statistics(UNIVERSITY OF M'SILA, 2025-06-30) DJEMAI Boutheyna; ENCA/ Mezache AmarIn this thesis, a comprehensive study of radar systems and their basic concepts are presented firstly. Next, the problem of interference faced by radars and its impact on detection accuracy is also evoked. Then, radar clutter types and targets fluctuating models are also highlighted. Research works concerned in this thesis deal with the improvements of high resolution sea clutter modeling and CFAR detection in presence of interfering targets. The first research problem focuses on the use of mixture generalized gamma (GG) distribution named Trimodal GG distribution in order to provide an accurate model that reflects the description characteristics of the majority of sea clutter scenarios. Comparison purposes are conducted against GG, mixture of two GG, K+noise and CG-LNT+noise (Compound Gaussian Log-Normal Texture) models. Parameters estimation is obtained by the well known LSA (Least Squares Approximation) approach. The modeling results show that the proposed Trimodal GG model is able to fit most scenes of IPIX (Intelligent PIxel X-band radar) real data. The second research problem is to enhance CFAR (Constant False Alarm Rate) properties in presence of homogeneous and heterogeneous CG-LNT sea clutter. To achieve this, existing WHWH (Weber Haykin-Weber Haykin))-CFAR and OS (Order Statistic)-CFAR algorithms are combined by means of a general test statistic. This detector is labeled WHWHOS-CFAR and is compared with available logt-, OS-, WH-, WHOS- and WHWH-CFAR detectors. Through, simulated and IPIX real data, it is shown that the proposed WHWHOS-CFAR detector exhibits the best CFAR properties and provides a worst CFAR loss only for a special value of the standard deviation clutter parameter.Item Open Access Sign Language Recognition System Case Study : Algerian signs(University of M’sila, 2025-06-30) Hachemi Manar Zahrat ELOla; Chennafi Karima; ENCA/ BRIK MouradConsidering that communication is essential for human connection, the deaf community faces unique obstacles. Therefore, sign language is the best alternative for overcoming these communication barriers, as it is considered the most effective means of communication, involving many hand movements. However, sign language is often misunderstood by those not part of the deaf community, necessitating the use of interpreters. This has led the community to develop techniques to facilitate interpretation tasks. Despite progress in deep learning, there is still limited research on recognizing and translating Algerian Arabic sign language. This lack of research has prompted us to focus specifically on advancing studies in Algerian Arabic sign language. This thesis introduces improved methodologies to construct a comprehensive framework for processing, translating, and generating Algerian Arabic sign language from input videos. We begin by utilizing the Mediapipe library for identifying human body parts. Then, for sign language recognition, particularly in Arabic, we employed three distinct models: Convolutional Neural Networks (CNN), 63 Long Short-Term Memory (LSTM), and a hybrid CNN-LSTM approach. Using the ArabSign-A dataset, we adapted it to focus on individual words, achieving an accuracy of 95.23% for the CNN model, 88.09% for the LSTM model, and 96.66% for the hybrid model. A comparative analysis was conducted to evaluate our methodology, demonstrating superior discrimination between static signs compared to prior researcItem Open Access AUTOMATIC KEYWORD DETECTION AND CRIME PREDICTION FROM PHONE CALL ANALYSIS (SEERIUM)(University of M’sila, 2025-06-30) HAROUN KHEIR; YASSINE ZELLAGUI; ENCA/ HERIZI AbdelghafourTraditional voice recognition systems predominantly rely on anatomical features such as vocal tract geometry and articulatory patterns. This study introduces an alternative approach centered on the acoustic characteristics of speech, with an emphasis on frequency-domain features. The proposed system integrates spectral analysis, Mel-Frequency Cepstral Coefficients (MFCCs), and deep learning architectures to perform three tasks: keyword detection, gender classification, and speaker identification. Using a custom dataset comprising real-world, noise-contaminated voice recordings, the model achieved an accuracy of 94% for keyword detection, 80.5% for gender classification, and 71% for speaker identification. These results underscore the robustness of frequency-based features in non-ideal conditions and highlight their applicability in privacy-sensitive and locally processed voice recognition systems. Future work will explore advanced neural architectures and signal enhancement techniques to further improve performance across diverse environments. A web-based platform was also developed to allow users to test the system via voice uploads and receive immediate analysis results without needing MATLAB.Item Open Access Study of quantum well solar cells based on CuIn1-xGaxSe2 absorbers using SCAPS-1D simulator(UNIVERSITY OF M’SILA, 2025-06-30) SALAMANI Maram safàa; BEDDIAR Abdrrahmen; ENCA/BOUCHAMA IdrisThis thesis presents a comprehensive numerical study of Cu(In₁₋ₓGaₓ)Se₂ (CIGS)-based solar cells. It begins with a detailed examination of the structure, composition, and advantages of the CIGS material for photovoltaic applications. Using the SCAPS-1D simulator, several simulations are performed to assess the influence of key physical parameters—such as gallium content, absorber thickness, acceptor concentration, and defect density—on solar cell performance. The thesis also explores the integration of quantum well (QW) structures within the CIGS absorber. Numerical modeling of CIGS/QW devices evaluates the effect of quantum well thickness, defect levels, and well depth on the photovoltaic performance. The results provide valuable insights into optimizing CIGS-based solar cells, particularly through quantum engineering approaches.