Génie Mécaniques

Permanent URI for this communityhttps://dspace.univ-boumerdes.dz/handle/123456789/67

Browse

Author: search.filters.author.Benazzouz, Djamel(Directeur de thèse)Has files: Yes

Search Results

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Mechatronic systems diagnostics using signal processing based techniques
    (Université M'Hamed Bougara Boumerdès : Faculté de Technologie, 2026) Damou, Ali; Benazzouz, Djamel(Directeur de thèse); Benazzouz, Djamel
    Industry 4.0, or the factory of the future, is advancing faster than ever, with intensive competition and a more challenging industry. In this competitive context, rotating machinery, particularly gearbox systems, plays a crucial role in transmitting power and altering speed and torque; they are essential for many industrial processes. As one of the key mechanical parts in gearbox, bearings are often the heart of a wide range of industrial mechanisms, they are regarded as essential elements in various applications, including wind turbines, helicopters, robots and aerospace systems. Consequently, a sudden malfunction can significantly impair system performance and may even result in catastrophic failure or total breakdown. Ensuring their reliable operation through efficient maintenance and fault detection strategies has become essential to prevent unexpected failures and system downtime. Predictive maintenance strategies based on condition monitoring have garnered substantial attention in recent years. Condition-Based Maintenance (CBM) plays a pivotal role in maintaining optimal system performance by predicting failures and minimizing unnecessary preventive interventions, thus reducing maintenance costs. Various CBM techniques have been developed to assess gearbox health, including acoustic emission, thermal monitoring, chemical analysis, current signature analysis, and vibration analysis. Among these, vibration-based condition monitoring has received particular focus, as the vibration signals generated by gearboxes contain valuable diagnostic information about their health state. This dissertation focuses on bearing fault diagnosis in gearbox systems, emphasizing the importance of predictive maintenance and early fault detection. However, these signals are inherently non-stationary, with defect signatures often masked by noise and irrelevant components, especially in early-stage faults. Moreover, the presence of strong non-Gaussian noise often conceals fault frequencies in the spectrum. Additionally, when multiple bearings within the system share identical dimensions, they exhibit similar characteristic fault frequencies. In such cases, even ifafaultfrequencyis detected in the vibration spectrum, it becomes challenging to identify which specific bearing is the source of the fault, complicating fault localization. This thesis also deals with this issue by developing a new automatic approach to detect, identify, and classify several bearing defects. The intelligent method is a combination of Wavelet Packet Transform (WPT) and machine learning techniques. WPT was employed to decompose vibration signals into discrete frequency bands; relevant features were extracted from each sub-band in the time domain, enabling the capturing of distinct fault characteristics across various frequency ranges sensitive to different fault conditions. These features served as inputs to machine learning classifiers, facilitating the accurate identification and classification of defective bearings. A dynamic simulation of a single stage spur gearbox was employed to generate vibration data under both healthy conditions and three distinct bearing defects, demonstrating the capability of the proposed approach to accurately detect and identify bearing defects across all cases
  • No Thumbnail Available
    Item
    Electrical systems faults diagnosis based on thermography and machine learning techniques
    (Université M'Hamed Bougara Boumerdès : Faculté de Technologie, 2025) Mahami, Amine; Benazzouz, Djamel(Directeur de thèse)
    The goal of using AI-driven conditional monitoring in electrical devices is to monitor and trace the beginning and development of deterioration prior to a failure. This degradation eventually results in a system malfunction that impacts the availability of the whole system. Early identification allows for a planned shutdown, averting catastrophic failure and guaranteeing more cost-effective and dependable operation. This study is divided into two major parts: the first part deals with the identification and categorization of faults in induction motors, and the second part deals with the detection and classification of faults in transformers. In machine health management, condition monitoring and problem diagnostics of electrical machines are important study areas. Using infrared thermography method (IRT), a new noncontact and nonintrusive experimental framework is used in the first portion of this thesis to monitor and diagnose defects in a three-phase induction motor. Using IRT to obtain a thermograph of the target machine is the first step in the process. The Speeded-Up Robust Features (SURF) detector and descriptor are then used to extract fault features from the IRT images using the bag-of-visual-word (BoVW) technique. Then, a group learning method known as Extremely Randomized Tree (ERT) is applied to automatically detect different types of induction motor defect patterns. Based on experimental IRT images, the efficacy of the suggested method is evaluated, showcasing its potential as a potent diagnostic tool with superior classification accuracy and stability over alternative approaches. The second part of the thesis presents an experimental framework that uses infrared thermography (IRT) to monitor and diagnose transformer defects in a non-intrusive and non-contact manner. Using IRT to obtain a thermograph of the intended machine is the first step in the process. GIST features are then taken from the database's reference image and every other image. Finally, a machine learning technique known as Support Vector Machine (SVM) is used to automatically identify different fault patterns in the transformer. Based on experimental IRT images and diagnostic results, the efficacy and capacity of the proposed method are assessed, demonstrating its potential as a potent diagnostic tool with high classification accuracy and stability. This method improves operational reliability by facilitating the early identification and detection of transformer failures
  • No Thumbnail Available
    Item
    Analysis of bearings defaults using machine learning techniques
    (Université M'Hamed Bougara Boumerdès : Faculté de Technologie, 2025) Moussaoui, Imane; Benazzouz, Djamel(Directeur de thèse)
    Rotating machines have become ubiquitous in contemporary industries, playing a pivotal role in various applications. The consequences of defects in these machines extend beyond mere technical issues, potentially leading to substantial economic losses and posing a significant threat to human safety. Operators often grapple with the intricacies of troubleshooting these complex systems, where a single mistake can have catastrophic consequences. One of the most critical elements in these machines is the bearings. Consequently, numerous researchers have dedicated their time and efforts to addressing this matter. While extensive studies have been conducted in this field, a common limitation is the focus on constant-speed scenarios. In reality, rotating machines typically operate under non-stationary conditions, making constant-speed techniques largely theoretical. This thesis is divided into two essential parts. The first part addresses the challenges of diagnostic resolution under time-varying conditions. Given the dynamic nature of the working environment, understanding and mitigating faults in non-stationary conditions is imperative for practical applications. Our method aims to tackle the diagnostic issue under time-varying conditions. The technique was tested on a bearing database collected under time-varying conditions, containing three types of faults. Vibrational signals are initially processed using the Empirical Wavelet Transform (EWT) to extract AM-FM modes. Subsequently, a list of features is extracted from these modes. For feature selection, the Clan-Based Cultural Algorithm (CCA) is employed, and model training utilizes the Random Forest algorithm. The results demonstrate the robustness of the diagnostic process despite varying conditions. The second part focuses on feature selection, which plays a crucial role in controlling the quality of the diagnostic system and reducing misleading factors. This area of research is increasingly attracting attention, with numerous methods developed. However, many of these techniques require in-depth domain knowledge, particularly concerning parameter tuning and result interpretation. In this work, we introduce a robust technique based on standard deviation and Random Forest methods for sequential feature selection. The method was tested on three different bearing databases, including time-varying conditions, and three signal decomposition techniques (EWT, EMD, and MODWPT). It provided promising results in terms of both quality and quantity, being user-friendly and not demanding extensive knowledge in the optimization field
  • No Thumbnail Available
    Item
    Energy-aware USVs path planning
    (Université M'Hamed Bougara : Faculté de Technologie, 2022) Ouelmokhtar, Hand; Benazzouz, Djamel(Directeur de thèse)
    Unmanned Surface Vehicles (USV) are an innovative solution for various maritime applications such as marine navigation, rescue, environmental monitoring and surveillance, etc. USVs offer the advantage to operate in hostile or dangerous environments where humans cannot safely or not at all perform. In general, USVs operate in harsh environmental conditions that require accuracy, reliability and autonomy. To meet these critical requirements, the focus on USVs and their applications is gradually performed. One of the most important problems to be solved is that of trajectory planning. In order to execute the planned tasks, the USVs must operate in an autonomous way and manage their resources optimally in order to minimize human interventions. Thus, performance and autonomy criteria are very important to consider when executing any type of task. In this thesis, we address the general problem of maritime surveillance using a USV equipped with an on-board LiDAR (Light Detection and Ranging) that allows remote coverage of distant points. The objectives are to cover the maximum area with lowest energy cost while avoiding collisions with obstacles. To solve this problem, we used two optimization approaches: • The first one consists in using heuristic methods based on multi-objective evolutionary algorithms. In this case, two algorithms are used and compared. One consists of a local search method known as Pareto Archived Evolution Strategy (PAES). Other consists of a population-based search algorithm called Non-Dominated Genetic Sorting Algorithm II (NSGA-II). • A novel method is proposed to improve the performance of evolutionary algorithms when solving path planning problems by reducing the size of chromosomes. • The second approach isbased on the exact method using a Mixed Integer Programming (MIP) model with two objective functions inspired by both the Covering Salesman Problem (CSP) and the Travelling Salesman Problem with Profit (TSPP).
  • No Thumbnail Available
    Item
    Contribution au diagnostic des systèmes par bond graph
    (Université M'Hamed Bougara : Faculté de Technologie, 2021) Termeche, Adel; Benazzouz, Djamel(Directeur de thèse)
    La détection et l'isolation des défauts (FDI) est une tâche essentielle qui permet d’éviter les conséquences des pannes sur les performances du système. Le Bond Graph, en tant qu'outil de modélisation, fournit à travers ses propriétés structurelles et causales, une génération automatique de Relations de Redondance Analytique (RRA). Ces relations sont utilisées pour les applications de diagnostic, elles constituent les contraintes mathématiques qui permettent de vérifier la cohérence entre les mesures du processus et le comportement du système de référence représenté par le modèle. L'approche de diagnostic RRA classique permet à la fois de détecter et d'isoler le composant défectueux du système. Dans ce travail, l'objectif principal est d'augmenter le nombre de défauts isolables en augmentant le nombre de RRA, en utilisant la sortie du modèle de graphe de liaison avec la sortie mesurée du système réel. La représentation Bond Graph sous forme de transformation linéaire fractionnelle (LFT) a été exploitée pour l’intégration de la fonction linéaire, cette dernière est utilisée pour l’amélioration de la détection des défauts en présence des erreurs de mesures. L'intérêt innovant de ce travail est que le nombre des défauts isolés peut être amélioré sans l'ajout de capteurs supplémentaires. Suite à la discussion générale de la méthode proposée, un sous-système robotique (traction d'un robot mobile omnidirectionnel) est envisagé pour valider la procédure proposée. Deux scénarii défectueux sont ensuite présentés et discutés en utilisant à la fois l'approche classique et l'approche proposée. La méthode proposée est capable d'isoler 3 défauts qui ne peuvent pas être isolées avec l’approche de RRA classique.