The diesel engine in heavy-duty trucks is a complex system with many components working together, and a malfunction in any of these components can impact engine performance and result in increased emissions. Fault detection and diagnosis have therefore become essential in modern vehicles, ensuring optimal performance and compliance with progressively stricter legal requirements. One of the most common faults in a diesel engineis faulty injectors, which can lead to fluctuations in the amount of fuel injected. Detecting these issues is crucial, prompting a growing interest in exploring additional signals beyond the currently used signal to enhance the performance and robustness of diagnosing this fault. In this work, an investigation was conducted to identify signals that correlate with faulty injectors causing over- and underfueling. It was found that the NOx, O2, and exhaust pressure signals are sensitive to this fault and could potentially serve as additional diagnostic signals. With these signals, two different diagnostic methods were evaluated to assess their effectiveness in detecting injector faults. The methods evaluated were data-driven residuals and Random Forest classifier. The data-driven residuals, when combined with the CUSUM algorithm, demonstrated promising results in detecting faulty injectors. The O2 signal proved effective in identifying both fault instances, while NOx and exhaust pressure were more effective at detecting overfueling. The Random Forest classifier also showed good performance in detecting both over- and underfueling. However, it was observed that using a classifier requires more extensive data preprocessing. Two preprocessing methods were employed: integrating previous measurements and calculating statistical measures over a defined time span. Both methods showed promising results, with the latter proving to be the better choice. Additionally, the generalization capabilities of these methods across different operating conditions were evaluated. It was demonstrated thatthe data-driven residuals yielded better results compared to the classifier, which requiredtraining on new cases to perform effectively.

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.

Les travaux développés durant cette thèse concernent la détection et l'identification des défauts simples et multiples d'ouverture des transistors dans un convertisseur statique associé à une machine asynchrone. Pour aborder cette problématique, nous avons commencé par l'analyse des potentialités, des faiblesses et des incertitudes des techniques qui ont initiés notre démarche. Ensuite, nous avons présenté deux méthodologies permettant d'analyser les performances du moteur asynchrone en présence des défauts dans une ou plusieurs cellules de commutation. Cette étude préliminaire nous a permis ainsi de proposer deux nouvelles stratégies de diagnostic sans référence basées sur l'approche signal. Les signaux électriques (courants ou tensions) disponibles à la sortie du convertisseur statique sont utilisés pour alimenter le processus de diagnostic. La première stratégie retenue est basée sur l'analyse qualitative des tensions de sortie entre phases du convertisseur et des signaux de commande appliqués aux transistors pendant les instants de commutation. Grâce à une représentation instantanée de ces grandeurs, à l'échelle de la période de découpage, nous avons pu mettre en évidence des caractéristiques favorables à la détection des défauts simples et multiples d'ouverture des transistors. L'implémentation pratique de cette première approche a été réalisée au moyen d'une technologie analogique permettant ainsi de minimiser le temps de retard à la détection jusqu'à quelques dizaines de microsecondes. / The main goal of this thesis concerns the detection and identification of simple and multiple open-circuit faults in voltage source inverters (VSIs)-fed induction motor drives. In first step, the potentialities, the weaknesses as well as the uncertainties of the previously published works have been discussed. The second step was dedicated to the study of the inverter faults impact on the induction motor. For this purpose, we have proposed two methodologies permitting the characterization of the electromagnetic torque behaviour as well as the electric variables of the induction motor under the open- and short-circuit faults. These preliminary studies allowed to propose two novel signal-based approaches for open-circuit fault diagnosis in voltage source inverter. The measured outputs inverter voltages and currents have been used as the input quantities for the fault detection and identification (FDI) process. The first approach consists in analyzing the pulse-width modulation (PWM) switching signals and the line-to-line voltage levels during the switching times, under both healthy and faulty operating conditions. For this purpose, we have adopted an instantaneous representation of these variables, which permits their analysis over one switching period. The fault diagnosis scheme is achieved using simple analog device. This circuit allows an accurate single and multiple faults diagnosis, and a minimization of the fault detection time which becomes about a few tens of microseconds.

O objetivo principal desta tese consiste no desenvolvimento de ferramentas matemáticas e computacionais dedicadas a um sistema de diagnóstico de barras quebradas no rotor de Motores de Indução Trifásicos. O sistema proposto é baseado em um método matemático de decomposição de sinais elétricos, denominado de Decomposição em Componentes Ortogonais, e ferramentas de aprendizagem de máquinas. Como uma das principais contribuições desta pesquisa, realizou-se um aprofundamento do entendimento da técnica de Decomposição em Componentes Ortogonais e de sua aplicabilidade como ferramenta de processamento de sinais para sistemas elétricos e eletromecânicos. Redes Neurais Artificiais e Support Vector Machines, tanto para classificação multi-classes quanto para detecção de novidades, foram configurados para receber índices advindos do processamento de sinais elétricos de motores, e a partir deles, identificar os padrões normais e os padrões com falhas. Além disso, a severidade da falha também é diagnosticada, a qual é representada pelo número de barras quebradas no rotor. Para a avaliação da metodologia, considerou-se o acionamento de motores de indução pela tensão de alimentação da rede e por inversores de frequência, operando sob diversas condições de torque de carga. Os resultados alcançados demonstram a eficácia das ferramentas matemáticas e computacionais desenvolvidas para o sistema de diagnóstico, sendo que os índices criados se mostraram altamente correlacionados com o fenômeno da falha. Mais especificamente, foi possível criar índices monotônicos com a severidade da falha e com baixa variabilidade, demonstrando-se que as ferramentas são eficientes extratores de características. / This doctoral thesis consists of the development of mathematical and computational tools dedicated to a diagnostic system for broken rotor bars in Three Phase Induction Motors. The proposed system is based on a mathematical method for decomposing electrical signals, named the Orthogonal Components Decomposition, and machine learning tools. As one of the main contributions of this research, an in-depth investigation of the decomposition technique and its applicability as a signal processing tool for electrical and electromechanical systems was carried-out. Artificial Neural Networks and Support Vector Machines for multi-class classification and novelty detection were configured to receive indices derived from the processing of electrical signals and then identify normal motors and faulty motors. In addition, the fault severity is also diagnosed, which is represented by the number of broken rotor bars. Experimental data was tested in order to evaluate the proposed method. Signals were obtained from induction motors operating with different torque levels and driven either directly by the grid or by frequency inverters. The results demonstrate the effectiveness of the mathematical and computational tools developed for the diagnostic system since the indices created are highly correlated with the fault phenomenon. More specifically, it was possible to create monotonic indices with the fault severity and with low variability, what supports that the solution is an efficient fault-specific feature extractor.

Nos sistemas de distribuição de alta e média tensão tem-se aumentado a utilização de linhas de distribuição de energia subterrâneas ou cabos subterrâneos. A ocorrência de faltas nas linhas afeta negativamente a qualidade da energia e o correto funcionamento da rede. O processo que leva a uma falta nos cabos é gradual e está caracterizado por uma série de subciclos de faltas incipientes associadas a uma tensão de arco. Estas, muitas vezes, passam despercebidas e, eventualmente, resultam em uma falta permanente. Os métodos clássicos de localização de faltas como as metodologias baseadas no cálculo da impedância aparente, as baseadas na inteligência artificial e as baseadas nas ondas viajantes são, habitualmente, aplicadas ao sistema depois de uma falta permanente acontecer e precisam de um ou mais ciclos do sinal para entregar uma resposta razoável. No entanto, as faltas nos cabos são um processo gradual, de curta duração (entre ¼ e ½ ciclo do sinal) e seria desejável localizar a falta incipiente antes de tornar-se permanente. Nesse contexto, esta pesquisa aborda o problema de localização de faltas incipientes. Assim, nesta tese propõe-se uma nova técnica de localização de faltas incipientes usando medições em um terminal, no domínio do tempo e que utiliza componentes de fase. Desta forma, são desenvolvidas duas novas formulações do modelo elétrico do sistema de distribuição com cabos subterrâneos durante uma falta incipiente. A abordagem proposta considera simultaneamente na sua formulação características da falta incipiente e dos sistemas de distribuição de energia, como a tensão de arco, o modelo Π nominal de parâmetros concentrados do cabo subterrâneo, o desequilíbrio do sistema e a condição da carga. A estimativa da distância da falta, junto com os parâmetros da falta incipiente, é obtida a partir da solução de um sistema sobredeterminado de equações lineares pela aplicação do método de mínimos quadrados ponderados não negativos. As formulações propostas permitem estimar a distância da falta em termos da reatância da linha até a falta. Além disso, é proposto um processo de compensação de corrente para estimar a corrente de falta; é aplicado um pré-processamento dos dados de entrada para suavizar o efeito do ruído que pode conter o sinal e, é aplicado um pós-processamento dos resultados para refinar e entregar a melhor estimativa obtida durante o processo de localização da falta. O desempenho da técnica proposta é avaliado mediante estudos de casos simulados em um circuito real de distribuição no Alternative Transients Program (ATP/EMTP) considerando análises de sensibilidade e comparativa. Também, o modelo da falta incipiente foi programado utilizando a ferramenta de MODELS do ATP/EMTP. Os resultados obtidos, considerando faltas incipientes simuladas que avaliam a influência da variação da magnitude de tensão de arco, do ruído aleatório inserido na tensão de arco, da distância da falta, da taxa de amostragem, do carregamento do sistema, do modelo de tensão de arco e de incertezas nas medições, indicam claramente que a abordagem proposta possui validade como técnica de localização de faltas incipientes, apresentando erros médios globais de 1,60% e 0,93%, respectivamente para cada formulação proposta. / The use of underground power distribution lines or underground cables in the high and medium voltage distribution systems has increased dramatically in recent years. The fault occurrence in the distribution lines negatively affects the power quality and the correct network operation. The process which leads to a fault in underground cables is gradual and is characterized by a series of sub-cycles of incipient faults associated with an arc voltage. These often are unnoticed and, eventually, results in a permanent fault. Classical fault localization methods such as the based-impedance, the based on artificial intelligent and the based on traveling waves are, usually, applied to the system after a permanent fault occurrence and need one or more signal cycles for providing a reasonable response. However, the faults in cables are a gradual process, with short duration (between ¼ to ½ of signal cycle) and would be desirable to locate the fault before this becomes a permanent fault. In this context, this research approaches the incipient faults location problem. Thus, in this thesis is proposed a new incipient fault location technique using single-end terminal measurement, in time-domain and employing phase components. In this way, two new formulations of the electrical model of the distribution system with underground cables during an incipient fault are developed. The proposed approach considers simultaneously in its formulation, incipient fault type and power distribution systems characteristics as arc voltage, unbalanced operation, load conditions and complete line model. The fault distance estimation, together with the incipient fault parameters, it is obtained from the solution of an overdetermined linear system of equations by the application of the non-negative weighted least squares estimator method. The proposed formulations allow estimating the fault distance in terms of the line reactance up to the fault. In addition, a load current compensation strategy is proposed to reduce its effect in the fault current estimation; an input data pre-processing is applied to smooth out the noise effect and a post-processing of the results is performed for estimation refinement and to provide the best estimate obtained during the fault location process. The proposed technique performance is evaluated through simulated cases studies in a real-life distribution network with underground cable data in the Alternative Transients Program (ATP/EMTP) considering sensitivity and comparative analyzes. Also, the fault model was programmed using the MODELS tool of ATP/EMTP. The obtained results, considering simulated incipient faults, which evaluate the influence of variations in the arc voltage magnitude, random noise percentage inserted in the arc voltage, fault distance, sampling rate, load dynamics, the arc voltage model and uncertainties in measurements, indicate clearly that the proposed approach is valid as incipient faults location technique, showing overall average errors of 1,60% and 0,93%, respectively for each proposed formulation.

Cette thèse s’inscrit dans le domaine de l'intelligence ambiante (Ambient Intelligence - AmI). Les systèmes AmI sont des systèmes interactifs composés de plusieurs éléments hétérogènes. Principalement : les capteurs et les effecteurs.D'un point de vue fonctionnel, l'objectif des systèmes AmI est d'activer certains effecteurs, sur la base des mesures des capteurs. Toutefois, les capteurs et les effecteurs peuvent subir des défaillances. Notre motivation dans cette thèse est de munir les systèmes AmI de capacités d'auto-détection des pannes.Les ressources physiques ne sont pas nécessairement connues au moment de la conception, mais elles sont plutôt découvertes dynamiquement lors de l'exécution. Il est donc impossible d’appliquer les techniques classiques pour prédéterminer des boucles de régulation ad-hoc.Nous proposons une nouvelle approche où la stratégie de détection de défaillances est déterminée dynamiquement lors de l'exécution. Pour cela, les couplages entre capteurs et effecteurs sont déduits automatiquement lors de l’exécution. Ceci est rendu possible par la modélisation des caractéristiques des capteurs, des effecteurs, ainsi que des phénomènes physiques (que nous appelons effets) qui sont attendus dans l'environnement ambiant suite à une action d’un effecteur. Ces effets sont utilisés en run-time pour lier les effecteurs (produisant les effets) avec les capteurs correspondants (détectant ces effets). Nous introduisons une plateforme de détection des pannes qui génère à l’exécution un modèle de prédiction des valeurs attendues sur les capteurs. Ce modèle, de nature hétérogène (il mêle flots de données et automates finis) est exécuté par un outil adapté (ModHel’X) de façon à fournir les valeurs attendues à chaque instant. Notre plateforme compare alors ces valeurs avec les valeurs réellement mesurées de façon à détecter les défaillances. / This thesis takes place in the field of Ambient Intelligence (AmI). AmI Systems are interactive systems composed of many heterogeneous components. From a hardware perspective these components can be divided into two main classes: sensors, using which the system observes its surroundings, and actuators, through which the system acts upon its surroundings in order to execute specific tasks.From a functional point of view, the goal of AmI Systems is to activate some actuators, based on data provided by some sensors. However, sensors and actuators may suffer failures. Our motivation in this thesis is to equip ambient systems with self fault detection capabilities. One of the particularities of AmI systems is that instances of physical resources (mainly sensors and actuators) are not necessarily known at design time; instead they are dynamically discovered at run-time. In consequence, one could not apply classical control theory to pre-determine closed control loops using the available sensors. We propose an approach in which the fault detection and diagnosis in AmI systems is dynamically done at run-time, while decoupling actuators and sensors at design time. We introduce a Fault Detection and Diagnosis framework modeling the generic characteristics of actuators and sensors, and the physical effects that are expected on the physical environment when a given action is performed by the system's actuators. These effects are then used at run-time to link actuators (that produce them) with the corresponding sensors (that detect them). Most importantly the mathematical model describing each effect allows the calculation of the expected readings of sensors. Comparing the predicted values with the actual values provided by sensors allows us to achieve fault-detection.

O objetivo principal desta tese consiste no desenvolvimento de ferramentas matemáticas e computacionais dedicadas a um sistema de diagnóstico de barras quebradas no rotor de Motores de Indução Trifásicos. O sistema proposto é baseado em um método matemático de decomposição de sinais elétricos, denominado de Decomposição em Componentes Ortogonais, e ferramentas de aprendizagem de máquinas. Como uma das principais contribuições desta pesquisa, realizou-se um aprofundamento do entendimento da técnica de Decomposição em Componentes Ortogonais e de sua aplicabilidade como ferramenta de processamento de sinais para sistemas elétricos e eletromecânicos. Redes Neurais Artificiais e Support Vector Machines, tanto para classificação multi-classes quanto para detecção de novidades, foram configurados para receber índices advindos do processamento de sinais elétricos de motores, e a partir deles, identificar os padrões normais e os padrões com falhas. Além disso, a severidade da falha também é diagnosticada, a qual é representada pelo número de barras quebradas no rotor. Para a avaliação da metodologia, considerou-se o acionamento de motores de indução pela tensão de alimentação da rede e por inversores de frequência, operando sob diversas condições de torque de carga. Os resultados alcançados demonstram a eficácia das ferramentas matemáticas e computacionais desenvolvidas para o sistema de diagnóstico, sendo que os índices criados se mostraram altamente correlacionados com o fenômeno da falha. Mais especificamente, foi possível criar índices monotônicos com a severidade da falha e com baixa variabilidade, demonstrando-se que as ferramentas são eficientes extratores de características. / This doctoral thesis consists of the development of mathematical and computational tools dedicated to a diagnostic system for broken rotor bars in Three Phase Induction Motors. The proposed system is based on a mathematical method for decomposing electrical signals, named the Orthogonal Components Decomposition, and machine learning tools. As one of the main contributions of this research, an in-depth investigation of the decomposition technique and its applicability as a signal processing tool for electrical and electromechanical systems was carried-out. Artificial Neural Networks and Support Vector Machines for multi-class classification and novelty detection were configured to receive indices derived from the processing of electrical signals and then identify normal motors and faulty motors. In addition, the fault severity is also diagnosed, which is represented by the number of broken rotor bars. Experimental data was tested in order to evaluate the proposed method. Signals were obtained from induction motors operating with different torque levels and driven either directly by the grid or by frequency inverters. The results demonstrate the effectiveness of the mathematical and computational tools developed for the diagnostic system since the indices created are highly correlated with the fault phenomenon. More specifically, it was possible to create monotonic indices with the fault severity and with low variability, what supports that the solution is an efficient fault-specific feature extractor.

Cette thèse porte sur le diagnostic des systèmes à événements discrets modélisés par des Réseaux de Petri labellisés (RdP-L). Les problèmes de diagnostic monolithique et de diagnostic modulaire sont abordés. Des contributions sont proposées pour résoudre les problèmes d'explosion combinatoire et de complexité de calcul. Dans le cadre de l'analyse de la diagnosticabilité monolithique, certaines règles de réduction sont proposées comme un complément pour la plupart des techniques existantes de l'analyse de la diagnosticabilité, qui simplifient le modèle RdP-L tout en préservant sa propriété de diagnosticabilité. Pour un RdP-L sauf et vivant, une nouvelle condition suffisante pour la diagnosticabilité est proposée. Pour un RdR-L borné et non bloquant après l'occurrence d'une faute, l'analyse à-la-volée est améliorée en utilisant la notion d'explications minimales qui permettent de compacter l'espace d'état ; et en utilisant des T-semiflots pour trouver rapidement un cycle indéterminé. Une analyse à-la-volée utilisant Verifier Nets (VN) est proposée pour analyser à la fois les RdP-L bornés et non-bornés, ce qui permet d'obtenir un compromis entre efficacité du calcul et limitation des explosions combinatoires. Dans le cadre de l'analyse de la diagnosticabilité modulaire, une nouvelle approche est proposée pour les RdP-Ls décomposés. Les règles de réduction, qui préservent la propriété de la diagnosticabilité modulaire, sont appliquées pour simplifier le modèle initial. La diagnosticabilité locale est analysée en construisant le VN et le Graphe d'Accessibilité Modifié (MAG) du modèle local. La diagnosticabilité modulaire est vérifiée en construisant la composition parallèle du MAG et des graphes d'accessibilités d'autres modules du système. La complexité de calcul est inférieure à celles des autre approches dans la littérature. D'autre part, l'explosion combinatoire est également réduite en utilisant la technique de ε-réduction / This thesis deals with fault diagnosis of discrete event systems modeled by labeled Petri nets (LPN). The monolithic diagnosability and modular diagnosability issues are addressed. The contributions are proposed to reduce the combinatorial explosion and the computational complexity problems. Regarding monolithic diagnosability analysis, some reduction rules are proposed as a complement for most diagnosability techniques, which simplify the LPN model and preserve the diagnosability property. For a safe and live LPN, a new sufficient condition for diagnosability is proposed. For a bounded LPN that does not deadlock after a fault, the on-the-fly diagnosability analysis is improved by using minimal explanations to compact the state space; and by using T-invariants, to find quickly an indeterminate cycle. An on-the-fly diagnosability analysis using Verifier Nets (VN) is proposed to analyze both bounded and unbounded LPN, which achieves a compromise between computation efficiency and combinatorial explosion limitation. Regarding modular diagnosability analysis, a new approach is proposed for decomposed LPNs model. Reduction rules, that preserve the modular diagnosability property, are applied to simplify the model. The local diagnosability is analyzed by building the VN and the Modified Reachability Graph (MRG) of the local model. The modular diagnosability is verified by building the parallel composition of the MRG and the reachability graphs of other modules of the system. We prove in this study that the computational complexity of our approach is lower than existing approaches of literature. The combinatorial explosion is also reduced by using the ε -reduction technique.

L’objectif de cette thèse est de proposer des stratégies de diagnostic dans le cas d'une commande en vitesse sans capteur mécanique (vitesse/position) d’une machine asynchrone triphasée en présence de défaut d'ouverture des transistors IGBT (Insulated Gate Bipolar Transistor) de l’onduleur. Une étude de l’impact de ces défauts sur les performances de ces structures sans capteur mécanique en termes de stabilité et de robustesse des observateurs en mode dégradé est présentée. Un observateur par mode glissant (Sliding Mode Observer) à base de modèle est développé et validé expérimentalement en vue de la commande sans capteur mécanique de la machine asynchrone triphasée. Les signaux issus de l’observateur (approche modèle) sont utilisés conjointement avec ceux mesurés (approche signale) pour former une approche hybride de diagnostic de défauts des transistors IGBT de l’onduleur. Un observateur par mode glissant d’ordre 2 à base d’un algorithme Super-Twisting est ensuite développé en vue d’améliorer la stabilité et d’assurer la continuité de fonctionnement du système en présence d'un défaut afin de pouvoir appliquer une stratégie de commande tolérante aux défauts dans les meilleures délais et conditions de fonctionnement. / The main goal of this thesis is to propose diagnostic strategies in the case of a sensorless speed control of a three-phase induction motor under an opened-switch or opened-phase fault. A qualitative analysis of the performances, in terms of stability and robustness, of sensorless control applied to the electrical drive in pre-fault and post-fault operation modes is presented. A model-based sliding mode observer is developed and experimentally validated for sensorless speed control of three-phase induction motor. The signals issued from the observer (model approach) as well as the measured ones (signal approach) are simultaneously used to form a hybrid approach for inverter open-switch fault detection and identification. A second-order sliding mode observer based on Super-Twisting algorithm (STA) is also developed to improve the stability and to ensure the continuity of operation of the electrical drive especially during transient states induced by the fault, permitting thus to apply the reconfiguration step without losing the control

Electric power distribution network is an important part of electrical power systems for delivering electricity to consumers. Electric power utilities worldwide are increasingly adopting the computer aided monitoring, control and management of electric power distribution systems to provide better services to the electrical consumers. Therefore, research and development activities worldwide are being carried out to automate the electric power distribution system. The power distribution system consists of a three-phase source supplying power through single-, two-, or three-phase distribution lines, switches, and transformers to a set of buses with a given load demand. In addition, unlike transmission systems, single-, two-, and three-phase sections exist in the network and single-, two-, and three-phase loads exist in the distribution networks. Further, most distribution systems are overhead systems, which are susceptible to faults caused by a variety of situations such as adverse weather conditions, equipment failure, traffic accidents, etc. When a fault occurs on a distribution line, it is very important for the utility to identify the fault location as quickly as possible for improving the service reliability. Hence, one of the crucial blocks in the operation of distribution system is that of fault detection and it’s location. The achievement of this objective depends on the success of the distribution automation system. The distribution automation system should be implemented quickly and accurately in order to isolate those affected branches from the healthy parts and to take alternative measures to restore normal power supply. Fault location in the distribution system is a difficult task due to its high complexity and difficulty caused by unique characteristics of the distribution system. These unique characteristics are discussed in the present work. In recent years, some techniques have been discussed for the location of faults, particularly in radial distribution systems. These methods use various algorithmic approaches, where the fault location is iteratively calculated by updating the fault current. Heuristic and Expert System approaches for locating fault in distribution system are also proposed which uses more measurements. Measurements are assumed to be available at the sending end of the faulty line segment, which are not true in reality as the measurements are only available at the substation and at limited nodes of the distribution networks through the use of remote terminal units. The emerging techniques of Artificial Intelligence (AI) can be a solution to this problem. Among the various AI based techniques like Expert systems, Fuzzy Set and ANN systems, the ANN approach for fault location is found to be encouraging. In this thesis, an ANN approaches with limited measurements are used to locate fault in long distribution networks with laterals. Initially the distribution system modeling (using actual a-b-c phase representation) for three-, two-, and single-phase laterals, three-, two-, and single- phase loads are described. Also an efficient three-phase load flow and short circuit analysis with loads are described which is used to simulate all types of fault conditions on distribution systems. In this work, function approximation (FA) is the main technique used and the classification techniques take a major supportive role to the FA problem. Fault location in distribution systems is explained as a FA problem, which is difficult to solve due to the various practical constraints particular to distribution systems. Incorporating classification techniques reduce this FA problem to simpler ones. The function that is approximated is the relation between the three-phase voltage and current measurements at the substation and at selected number of buses (inputs), and the line impedance of the fault points from the substation (outputs). This function is approximated by feed forward neural network (FFNN). Similarly for solving the classification problems such as fault type classification and source short circuit level classification, Radial Basis Probabilistic Neural Network (RBPNN) has been employed. The work presented in this thesis is the combinational use of FFNN and RBPNN for estimating the fault location. Levenberg Marquardt learning method, which is robust and fast, is used for training FFNN. A typical unbalanced 11-node test system, an IEEE 34 nodes test system and a practical 69- bus long distribution systems with different configurations are considered for the study. The results show that the proposed approaches of fault location gives accurate results in terms of estimated fault location. Practical situations in distribution systems such as unbalanced loading, three-, two-, and single- phase laterals, limited measurements available, all types of faults, a wide range of varying source short circuit levels, varying loading conditions, long feeders with multiple laterals and different network configurations are considered for the study. The result shows the feasibility of applying the proposed method in practical distribution system fault diagnosis.