Cette thèse est consacrée à l'étude de diagnostic de pannes pour les systèmes pile à combustible de type PEMFC. Le but est d'améliorer la fiabilité et la durabilité de la membrane électrolyte polymère afin de promouvoir la commercialisation de la technologie des piles à combustible. Les approches explorées dans cette thèse sont celles du diagnostic guidé par les données. Les techniques basées sur la reconnaissance de forme sont les plus utilisées. Dans ce travail, les variables considérées sont les tensions des cellules. Les résultats établis dans le cadre de la thèse peuvent être regroupés en trois contributions principales.La première contribution est constituée d'une étude comparative. Plus précisément, plusieurs méthodes sont explorées puis comparées en vue de déterminer une stratégie précise et offrant un coût de calcul optimal.La deuxième contribution concerne le diagnostic online sans connaissance complète des défauts au préalable. Il s'agit d'une technique adaptative qui permet d'appréhender l'apparition de nouveaux types de défauts. Cette technique est fondée sur la méthodologie SSM-SVM et les règles de détection et de localisation ont été améliorées pour répondre au problème du diagnostic en temps réel.La troisième contribution est obtenue à partir méthodologie fondée sur l'utilisation partielle de modèles dynamiques. Le principe de détection et localisation de défauts est fondé sur des techniques d'identification et sur la génération de résidus directement à partir des données d'exploitation.Toutes les stratégies proposées dans le cadre de la thèse ont été testées à travers des données expérimentales et validées sur un système embarqué. / Aiming at improving the reliability and durability of Polymer Electrolyte Membrane Fuel Cell (PEMFC) systems and promote the commercialization of fuel cell technologies, this thesis work is dedicated to the fault diagnosis study for PEMFC systems. Data-driven fault diagnosis is the main focus in this thesis. As a main branch of data-driven fault diagnosis, the methods based on pattern classification techniques are firstly studied. Taking individual fuel cell voltages as original diagnosis variables, several representative methodologies are investigated and compared from the perspective of online implementation.Specific to the defects of conventional classification based diagnosis methods, a novel diagnosis strategy is proposed. A new classifier named Sphere-Shaped Multi-class Support Vector Machine (SSM-SVM) and modified diagnostic rules are utilized to realize the novel fault recognition. While an incremental learning method is extended to achieve the online adaptation.Apart from the classification based diagnosis approach, a so-called partial model-based data-driven approach is introduced to handle PEMFC diagnosis in dynamic processes. With the aid of a subspace identification method (SIM), the model-based residual generation is designed directly from the normal and dynamic operating data. Then, fault detection and isolation are further realized by evaluating the generated residuals.The proposed diagnosis strategies have been verified using the experimental data which cover a set of representative faults and different PEMFC stacks. The preliminary online implementation results with an embedded system are also supplied.

Les travaux de recherches sur la commande, le diagnostic et la tolérance aux défauts appliqués aux drones deviennent de plus en plus populaires. Il est judicieux de concevoir des lois de commande qui garantissent la stabilité et les performances du drone, non seulement dans le cas nominal, mais également en présence de fortes perturbations et de défauts.Dans cette thèse, un nouvel algorithme bio-inspiré adapté pour la recherche de solutions dans des problèmes d’optimisation est développé. Cet algorithme est utilisé pour trouver les gains des différents contrôleurs conçus pour les drones. La commande par mode glissant est utilisée pour développer deux contrôleurs passifs tolérants aux défauts pour les quadrirotors: un contrôleur par mode glissant augmentée avec un intégrateur, et un contrôleur par mode glissant implémenté en cascade. Parce que les commandes passives ont une robustesse réduite, une commande active par mode glissant est développée. Pour traiter les défauts extrêmes, un contrôleur d’urgence basé sur la conversion du quadrirotor en trirotor est développé. Les commandes actives, passives, et le contrôleur d’urgences sont ensuite intégrés pour former un contrôleur tolérant aux défauts capable de gérer un grand nombre de défaillances tout en garantissant les ressources actionneur et en limitant la charge de calcul du processeur. Finalement, des contrôleurs tolérants aux défauts, actifs et passifs, basés sur des méthodes par mode glissant du premier et deuxième ordre sont développées pour les octorotors. La commande active utilise des méthodes d’allocation de contrôles pour redistribuer les efforts sur les actionneurs sains, réduisant ainsi l’effet du défaut. / Unmanned Aerial Vehicles (UAV) are more and more popular for their civil and military applications. Classical control laws usually show weaknesses in the presence of parameter uncertainties, environmental disturbances, and actuator and sensor faults. Therefore, it is judicious to design a control law capable of stabilizing the UAV not only in the fault-free nominal cases, but also in the presence of disturbances and faults. In this thesis, a new bio-inspired search algorithm called Ecological Systems Algorithm (ESA) suitable for engineering optimization problems is developed. The algorithm is used over the thesis to find optimal gains for the fault tolerant controllers. Sliding Mode Control theory is used to develop two Passive Fault Tolerant Controllers for quadrotor UAVs: Regular and Cascaded SMC. Because Passive Controllers handle a few numbers of faults, an Active Sliding Mode Fault Tolerant Controller using Kalman Filter is developed. To overcome severe faults and failures, an emergency controller based on the Quadrotor-to-Trirotor conversion maneuver is developed. The Controllers developed so far (Passive, Active, and emergency controllers) are then integrated to form the Integrated Fault Tolerant Controller (IFTC). The IFTC is a powerful controller that is able to handle a wide number of faults, and save actuator resources as well as processor computational effort. Finally, Passive and Active Fault Tolerant Controllers are designed for octorotor UAVs based on First Order and Second Order Sliding Mode Control. The AFTC uses Dynamic and Pseudo-Inverse Control Allocation methods to redistribute the control effort among healthy actuators reducing the effect of fault.

Permanent magnet synchronous machine drives are used more often. Although, synchronous machines drive also suffer from possible faults. This thesis is focused on the detection of the three-phase synchronous motor winding faults and the detection of the drive control loop sensors' faults. Firstly, a model of the faulty winding of the motor is presented. Effects of the inter-turn short fault were analyzed. The model was experimentally verified by fault emulation on the test bench with an industrial synchronous motor. Inter-turn short fault detection algorithms are summarized. Three existing conventional winding fault methods based on signal processing of the stator voltage and stator current residuals were verified. Three new winding fault detection methods were developed by the author. These methods use a modified motor model and the extended Kalman filter state estimator. Practical implementation of the algorithms on a microcontroller is described and experimental results show the performance of the presented algorithms in different scenarios on test bench measurements. Highly related motor control loop sensors fault detection algorithms are also described. These algorithms are complementary to winding fault algorithms. The decision mechanism integrates outputs of sensor and winding fault detection algorithms and provides an overall drive fault diagnosis concept.

Fault diagnosis can be used to early detect faults in a technical system, which means that workshop service can be planned before a component is fully degraded. Fault diagnosis helps with avoiding downtime, accidents and can be used to reduce emissions for certain applications. Traditionally, however, diagnosis systems have been designed using ad hoc methods and a lot of system knowledge. Model-based diagnosis is a systematic way of designing diagnosis systems that is modular and offers high performance. A model-based diagnosis system can be designed by making use of mathematical models that are otherwise used for simulation and control applications. A downside of model-based diagnosis is the modeling effort needed when no accurate models are available, which can take a large amount of time. This has motivated the use of data-driven diagnosis. Data-driven methods do not require as much system knowledge and modeling effort though they require large amounts of data and data from faults that can be hard to gather. Hybrid fault diagnosis methods combining models and training data can take advantage of both approaches decreasing the amount of time needed for modeling and does not require data from faults. In this thesis work a combined data-driven and model-based fault diagnosis system has been developed and evaluated for the exhaust treatment system in a heavy-duty diesel engine truck. The diagnosis system combines physical insights and neural networks to detect and isolate faults for the exhaust treatment system. This diagnosis system is compared with another system developed during this thesis using only model-based methods. Experiments have been done by using data from a heavy-duty truck from Scania. The results show the effectiveness of both methods in an industrial setting. It is shown how model-based approaches can be used to improve diagnostic performance. The hybrid method is showed to be an efficient way of developing a diagnosis system. Some downsides are highlighted such as the performance of the system developed using data-driven and model-based methods depending on the quality of the training data. Future work regarding the modularity and transferability of the hybrid method can be done for further evaluation.

This work is dedicated to faulty induction motors. These motors are often used in industrial applications thanks to their usability and their robustness. However, nowadays optimisation of production becomes so critical that the conceptual reliability of the motor is not sufficient anymore. Motor condition monitoring is expanding to serve maintenance planning and uptime maximisation. Moreover, the use of drive control sensors (namely stator current and voltage) can avoid the installation and maintenance of dedicated sensors for condition monitoring.<p><p>Many authors are working in this field but few approach the diagnosis from a detailed and clear physical understanding of the localised phenomena linked to the faults. Broken bars are known to modulate stator currents but it is shown in this work that it also changes machine saturation level in the neighbourhood of the bar. Furthermore, depending on the voltage level, this change in local saturation affects the amplitude and the phase of the modulation. This is of major importance as most diagnosis techniques use this feature to detect and quantify broken bars. For stator short-circuits, a high current is flowing in the short-circuited coil due to mutual coupling with the other windings and current spikes are flowing in the rotor bars as they pass in front of the short-circuited conductors. In the case of rotor eccentricities, the number of pole-pairs and the connection of these pole-pairs greatly affect the airgap flux density distribution as well as the repartition of the line currents in the different pole-pairs.<p><p>These conclusions are obtained through the use of time-stepping finite element models of the faulty motors. Moreover, circuit models of faulty machines are built based on the conclusions of previously explained fault analysis and on classical Park models. A common mathematical description is used which allows objective comparison of the models for representation of the machine behaviour and computing time.<p><p>The identifiability of the parameters of the models as well as methods for their identification are studied. Focus is set on the representation of the machine behaviour using these parameters more than the precise identification of the parameters. It is shown that some classical parameters can not be uniquely identified using only stator measurements.<p><p>Fault detection and identification using computationally cheap models are compared to advanced detection through motor stator current spectral analysis. This last approach allows faster detection and identification of the fault but leads to incorrect conclusions in low load conditions, in transient situations or in perturbed environments (i.e. fluctuating load torque and unideal supply). Efficient quantification of the fault can be obtained using detection techniques based on the comparison of the process to a model.<p><p>Finally, the work provides guidelines for motor supervision strategies depending on the context of motor utilisation. / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished

This thesis aims to use vibration and acoustic noise analysis to help a production line of a certain motor type to ensure good quality. Noise from the gearbox is sometimes present and the way it is detected is with a human listening to it. This type of error detection is subjective, and it is possible for human error to be present. Therefore, an automatic test that pass or fail the produced Brush Less Direct Current (BLDC)-motors is wanted. Two measurement setups were used. One was based on an accelerometer which was used for vibration measurements, and the other based on a microphone for acoustic sound measurements. The acquisition and analysis of the measurements were implemented using the data acquisition device, compactDAQ NI 9171, and the graphical programming software, NI LabVIEW. Two methods, i.e., power spectrum analysis and machine learning, were used for the analyzing of vibration and acoustic signals, and identifying faults in the gearbox. The first method based on the Fast Fourier transform (FFT) was used to the recorded sound from the BLDC-motor with the integrated planetary gearbox to identify the peaks of the sound signals. The source of the acoustic sound is from a faulty planet gear, in which a flank of a tooth had an indentation. Which could be measured and analyzed. It sounded like noise, which can be used as the indications of faults in gears. The second method was based on the BLDC-motors vibration characteristics and uses supervised machine learning to separate healthy motors from the faulty ones. Support Vector Machine (SVM) is the suggested machine learning algorithm and 23 different features are used. The best performing model was a Coarse Gaussian SVM, with an overall accuracy of 92.25 % on the validation data.

Heutzutage gewinnt die Simulation von Gebäuden und Anlagen zunehmend an Bedeutung, um die Betriebsweise der Anlagen zu diagnostizieren bzw. zu bewerten und den Energiebedarf vorherzusagen. Dabei hängt die erzielte Genauigkeit von dem Kompliziertheitsgrad des angewendeten Simulationsprogramms ab. Deshalb ist Modellbildung und -validierung ein sehr wichtiger Bestandteil eines Softwareentwicklungsprozesses, um die Zuverlässigkeit zu sichern. Am Institut für Thermodynamik und Technische Gebäudeausrüstung liegen zahlreiche Simulationsmodelle vor. Im Rahmen dieser vorliegenden Arbeit wurden weitere benötigte Modelle (hygrisches Verhalten der Wände (vereinfachtes Verfahren), Rippenrohrwärmeüberträger, Wärmeregenerator et al.) entwickelt und in das Programm TRNSYS eingefügt sowie die vorhandenen Modelle an ihre Genauigkeit angepasst. Insbesondere sind dies die Modelle für Splitsysteme bei stetiger und nichtstetiger Regelung mit der detaillierten Betrachtung des Anlagenverhaltens sowohl beim Voll- als auch beim Teillastbetrieb. Damit ist es erstmals gelungen, das gesamte Anlagensystem der Splittechnik ausführlich zu beschreiben. Um die analytische Validierung durchführen zu können, wurden die analytischen Modelle für eine Splitanlage bei stetiger und nichtstetiger Regelung unter den vordefinierten Randbedingungen entwickelt. Zur analytischen Validierung finden auch die vorhandenen Simulationsmodelle Anwendung, so dass sich die meisten Komponenten und das Simulationsprogramm TRNSYS verifizieren ließen. Diese Validierung erfolgte im Rahmen des IEA-SHC/HVAC BESTEST TASK 22. Da an diesem TASK verschiedene Forschungsinstitutionen mit jeweils unterschiedlichen Simulationsprogrammen teilnahmen, ergab sich die beste Möglichkeit, vergleichende Tests durchzuführen. Wenn dabei ein Programm signifikante Unterschiede zu den anderen liefert, liegt dies nicht immer an Programmfehlern. Aber kollektive Erfahrungen aus diesem TASK zeigen, dass bei Abweichungen meistens Fehler bzw. fragwürdige Algorithmen gefunden wurden. Nachdem das Simulationsprogramm TRNSYS validiert war, erfolgte die Erstellung eines Konzeptes zur Fehlererkennung und Diagnose der Regelstrategien von RLTA. Das Verfahren erlaubt sowohl die Beseitigung der möglichen Fehler in der Planungsphase beim Entwurf der Regelstrategien als auch den Test der vorhandenen Regelstrategien. Dies erhöht die Zuverlässigkeit und damit die Sicherheit beim Anlagenbetrieb. Schließlich dient das Verfahren als Werkzeug zur Optimierung der Betriebsweise von RLTA. Das Regelverhalten wurde anhand typischer Fälle vorgestellt und diskutiert. Mit Hilfe des Verfahrens zur Fehlererkennung und Diagnose der Betriebsweise von RLTA ließen sich vorhandene Regelstrategien testen und verbessern.

In the hope to increase the detection rate of faults in combined heat and power plant boilers thus lowering unplanned maintenance three machine learning models are constructed and evaluated. The algorithms; k-Nearest Neighbor, One-Class Support Vector Machine, and Auto-encoder have a proven track record in research for anomaly detection, but are relatively unexplored for industrial applications such as this one due to the difficulty in collecting non-artificial labeled data in the field.The baseline versions of the k-Nearest Neighbor and Auto-encoder performed very similarly. Nevertheless, the Auto-encoder was slightly better and reached an area under the precision-recall curve (AUPRC) of 0.966 and 0.615 on the trainingand test period, respectively. However, no sufficiently good results were reached with the One-Class Support Vector Machine. The Auto-encoder was made more sophisticated to see how much performance could be increased. It was found that the AUPRC could be increased to 0.987 and 0.801 on the trainingand test period, respectively. Additionally, the model was able to detect and generate one alarm for each incident period that occurred under the test period.The conclusion is that ML can successfully be utilized to detect faults at an earlier stage and potentially circumvent otherwise costly unplanned maintenance. Nevertheless, there is still a lot of room for improvements in the model and the collection of the data. / I hopp om att öka identifieringsgraden av störningar i kraftvärmepannor och därigenom minska oplanerat underhåll konstrueras och evalueras tre maskininlärningsmodeller.Algoritmerna; k-Nearest Neighbor, One-Class Support Vector Machine, och Autoencoder har bevisad framgång inom forskning av anomalidetektion, men är relativt outforskade för industriella applikationer som denna på grund av svårigheten att samla in icke-artificiell uppmärkt data inom området.Grundversionerna av k-Nearest Neighbor och Auto-encoder presterade nästan likvärdigt. Dock var Auto-encoder-modellen lite bättre och nådde ett AUPRC-värde av 0.966 respektive 0.615 på träningsoch testperioden. Inget tillräckligt bra resultat nåddes med One-Class Support Vector Machine. Auto-encoder-modellen gjordes mer sofistikerad för att se hur mycket prestandan kunde ökas. Det visade sig att AUPRC-värdet kunde ökas till 0.987 respektive 0.801 under träningsoch testperioden. Dessutom lyckades modellen identifiera och generera ett larm vardera för alla incidenter under testperioden. Slutsatsen är att ML framgångsrikt kan användas för att identifiera störningar iett tidigare skede och därigenom potentiellt kringgå i annat fall dyra oplanerade underhåll. Emellertid finns det fortfarande mycket utrymme för förbättringar av modellen samt inom insamlingen av data.

<p dir="ltr">Condition-based maintenance is a proactive maintenance strategy that prevents failures or diminished functionality in process systems through proper monitoring and management of process conditions. Despite being considered a mature maintenance management strategy in various industries, condition-based maintenance remains underutilized in pharmaceutical manufacturing. This situation needs to change, especially as the pharmaceutical industry continues to shift from batch to continuous manufacturing, where the implementation of CBM as a maintenance strategy assumes a greater importance.</p><p dir="ltr">This dissertation focused on addressing the challenges of implementing CBM in a continuous drug product manufacturing system. These challenges stem from the unique aspects of pharmaceutical drug product manufacturing, which includes the peculiar behavior of particulate materials and the evolutionary nature of pharmaceutical process development. The proposed solutions to address these challenges revolve around an innovative framework for the practical development of condition monitoring systems. Overall, this framework enables the incorporation of limited process knowledge in creating condition monitoring systems, which has the desired effect of empowering data-driven machine learning models.</p><p dir="ltr">A key feature of this framework is a formalized method to represent the process condition, which is usually vaguely defined in literature. This representation allows the proper mapping of preexisting condition monitoring systems, and the segmentation of the entire process condition model into smaller modules that have more manageable condition monitoring problems. Because this representation methodology is based on probabilistic graphical modelling, the smaller modules can then be holistically integrated via their probabilistic relationships, allowing the robust operation of the resulting condition monitoring system and the process it monitors.</p><p dir="ltr">Breaking down the process condition model into smaller segments is crucial for introducing novel fault detection capabilities, which enhances model prediction transparency and ensures prediction acceptance by a human operator. In this work, a methodology based on prediction probabilities was introduced for developing condition monitoring systems with novel fault detection capabilities. This approach relies on high-performing machine learning models capable of consistently classifying all the initially known conditions in the fault library with a high degree of certainty. Simplifying the condition monitoring problem through modularization facilitates this, as machine learning models tend to perform better on simpler systems. Performance indices were proposed to evaluate the novel fault detection capabilities of machine learning models, and a formal approach to managing novel faults was introduced.</p><p dir="ltr">Another benefit of modularization is the identification of condition monitoring blind spots. Applying it to the RC led to sensor development projects such as the virtual sensor for measuring granule flowability. This sensor concept was demonstrated successfully by using a data-driven model to predict granule flowability based on size and shape distribution measurements. With proper model selection and feature extraction guided by domain expertise, the resulting sensor achieved the best prediction performance reported in literature for granule flowability.</p><p dir="ltr">As a demonstration exercise in examining newly discovered faults, this work investigated a roll compaction phenomenon that is usually concealed from observation due to equipment design. This phenomenon results in the ribbon splitting along its thickness as it comes out of the rolls. In this work, important aspects of ribbon splitting were elucidated, particularly its predictability based on RC parameters and the composition of the powder blend used to form the ribbon. These findings have positive ramifications for the condition monitoring of the RC, as correspondence with industrial practitioners suggests that a split ribbon is desirable in some cases, despite being generally regarded as undesirable in the limited literature available on the subject.</p><p dir="ltr">Finally, this framework was primarily developed for the pharmaceutical dry granulation line, which consists of particle-based systems with a moderate level of complexity. However, it was also demonstrated to be feasible for the Tennessee Eastman Process (TEP), a more complex liquid-gas process system with a greater number of process faults, variables, and unit operations. Applying the framework resulted in machine learning models that yielded one of the best fault detection performances reported in literature for the TEP, while also introducing additional capabilities not yet normally reported in literature, such as fault diagnosis and novel fault detection.</p>

The increasing complexity of telecom networks, induced by the recent development of 5G, is a challenge for detecting faults in the telecom network. In addition to the structural complexity of telecommunication systems, data accessibility has become an issue both in terms of privacy and access cost. We propose a method relying on bi-level Federated Graph Neural Networks to identify anomalies in the telecom network while ensuring reduced communication costs as well as data privacy. Our method considers telecom data as a bi-level graph, where the highest level graph represents the interaction between sites, and each site is further expanded to its software (SW) performance behaviour graph. We developed and compared 4G/5G SW Fault Detection models under 3 settings: (1) Centralized Temporal Graph Neural Networks model: we propose a model to detect anomalies in 4G/5G telecom data. (2) Federated Temporal Graph Neural Networks model: we propose Federated Learning (FL) as a mechanism for privacy-aware training of models for fault detection. (3) Personalized Federated Temporal Graph Neural Networks model: we propose a novel aggregation technique, referred to as FedGraph, leveraging both a graph and the similarities between sites for aggregating the models and proposing models more personalized to each site’s behaviour. We compare the benefits of Federated Learning (FL) models (2) and (3) with centralized training (1) in terms of SW performance data modelling, anomaly detection, and communication cost. The evaluation includes both a scenario with normal functioning sites and a scenario where only a subset of sites exhibit faulty behaviour. The combination of SW execution graphs with GNNs has shown improved modelling performance and minor gains in centralized settings (1). In a normal network context, FL models (2) and (3) perform comparably to centralized training (CL), with slight improvements observed when using the personalized strategy (3). However, in abnormal network scenarios, Federated Learning falls short of achieving comparable detection performance to centralized training. This is due to the unintended learning of abnormal site behaviour, particularly when employing the personalized model (3). These findings highlight the importance of carefully assessing and selecting suitable FL strategies for anomaly detection and model training on telecom network data. / Den ökande komplexiteten i telenäten, som är en följd av den senaste utvecklingen av 5G, är en utmaning när det gäller att upptäcka fel i telenäten. Förutom den strukturella komplexiteten i telekommunikationssystem har datatillgänglighet blivit ett problem både när det gäller integritet och åtkomstkostnader. Vi föreslår en metod som bygger på Federated Graph Neural Networks på två nivåer för att identifiera avvikelser i telenätet och samtidigt säkerställa minskade kommunikationskostnader samt dataintegritet. Vår metod betraktar telekomdata som en graf på två nivåer, där grafen på den högsta nivån representerar interaktionen mellan webbplatser, och varje webbplats utvidgas ytterligare till sin graf för programvarans (SW) prestandabeteende. Vi utvecklade och jämförde 4G/5G SW-feldetekteringsmodeller under 3 inställningar: (1) Central Temporal Graph Neural Networks-modell: vi föreslår en modell för att upptäcka avvikelser i 4G/5G-telekomdata. (2) Federated Temporal Graph Neural Networks-modell: vi föreslår Federated Learning (FL) som en mekanism för integritetsmedveten utbildning av modeller för feldetektering. I motsats till centraliserad inlärning aggregeras lokalt tränade modeller på serversidan och skickas tillbaka till klienterna utan att data läcker ut mellan klienterna och servern, vilket säkerställer integritetsskyddande samarbetsutbildning. (3) Personaliserad Federated Temporal Graph Neural Networks-modell: vi föreslår en ny aggregeringsteknik, kallad FedGraph, som utnyttjar både en graf och likheterna mellan webbplatser för att aggregera modellerna. Vi jämför fördelarna med modellerna Federated Learning (FL) (2) och (3) med centraliserad utbildning (1) när det gäller datamodellering av SW-prestanda, anomalidetektering och kommunikationskostnader. Utvärderingen omfattar både ett scenario med normalt fungerande anläggningar och ett scenario där endast en delmängd av anläggningarna uppvisar felaktigt beteende. Kombinationen av SW-exekveringsgrafer med GNN har visat förbättrad modelleringsprestanda och mindre vinster i centraliserade inställningar (1). I en normal nätverkskontext presterar FL-modellerna (2) och (3) jämförbart med centraliserad träning (CL), med små förbättringar observerade när den personliga strategin används (3). I onormala nätverksscenarier kan Federated Learning dock inte uppnå jämförbar detekteringsprestanda med centraliserad träning. Detta beror på oavsiktlig inlärning av onormalt beteende på webbplatsen, särskilt när man använder den personliga modellen (3). Dessa resultat belyser vikten av att noggrant bedöma och välja lämpliga FL-strategier för anomalidetektering och modellträning på telekomnätdata.