With the deregulation of power industry and the market competition, stable and reliable power supply is a major concern of the independent system operator (ISO). Power-quality (PQ) study has become a more and more important subject lately. Harmonics, voltage swell, voltage sag, and power interruption could downgrade the service quality. In recent years, high speed railway (HSR) and massive rapid transit (MRT) system have been rapidly developed, with the applications of widespread semiconductor technologies in the auto-traction system. The harmonic distortion level worsens due to these increased uses of electronic equipment and non-linear loads. To ensure the PQ, power-quality disturbances (PQD) detection becomes important. A detection method with classification capability will be helpful for detecting disturbance locations and types. Electric machinery fault diagnosis is another issue of considerable attentions from utilities and customers. ISO need to provide a high quality service to retain their customers. Fault diagnosis of turbine-generator has a great effect on the benefit of power plants. The generator fault not only damages the generator itself, but also causes outages and loss of profits. With high-temperature, high-pressure and factors such as thermal fatigues, many components may go wrong, which will not only lead to great economic loss, but sometimes a threat to social security. Therefore, it is necessary to detect generator faults and take immediate actions to cut the loss. Besides, induction motor plays a major role in a power system. For saving cost, it is important to run periodical inspections to detect incipient faults inside the motor. Preventive techniques for early detection can find out the incipient faults and avoid outages. This dissertation developed various soft computing (SC) algorithms for detection including power-quality disturbances (PQD), turbine-generator fault diagnosis, and induction motor fault diagnosis. The proposed SC algorithms included support vector machine (SVM), grey clustering analysis (GCA), and probabilistic neural network (PNN). Integrating the proposed diagnostic procedure and existing monitoring instruments, a well-monitored power system will be constructed without extra devices. Finally, all the methods in the dissertation give reasonable and practical estimation method. Compared with conventional method, the test results showed a high accuracy, good robustness, and a faster processing performance.

Orientador: Aparecido Carlos Gonçalves / Resumo: Pares engrenados são elementos de transmissão de potência amplamente utilizados em máquinas e equipamentos, todavia as falhas catastróficas desses componentes são comuns e dispendiosas. A análise de vibrações está entre as técnicas de diagnóstico de defeitos incipientes utilizadas em manutenção preditiva, posto que a presença de uma falha altera o comportamento dinâmico do sistema e o estado de degradação pode ser detectado pelo monitoramento dos sinais de vibração. Na indústria atual, onde as aquisições de dados, tanto para controle de processos, quanto para o monitoramento das condições de integridade de equipamentos, são realizadas em tempo real, faz-se necessário o desenvolvimento de métodos que aumentem a confiabilidade das tomadas de decisões em relação à identificação, localização e prognóstico de falhas. O objetivo deste trabalho é desenvolver um modelo matemático de par de engrenagens que auxilie no monitoramento da condição e validar o modelo dinâmico com dados de vibração de um multiplicador de velocidades obtidos experimentalmente. Para tanto, foi elaborada uma metodologia baseada no modelo dinâmico de par engrenado com 6 graus de liberdade para simulação de sinais de vibração; nesse modelo, inclui-se erros geométricos no perfil do dente; de maneira analítica, simula-se uma a trinca do dente de uma das engrenagens que ocasiona a queda de rigidez em função do tempo; desenvolveu-se também um experimento com um multiplicador de velocidades; e, por fim, algumas técnic... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: Spur gears are transmission power elements widely used in machinery, however catastrophic failures of this components are just as common and onerous. Vibration analysis is a technique, in among of others, that can be used in diagnostics of incipient damages, common in predictive maintenance, because they change the dynamic behavior of the mechanical system, and the degradation state can be detected by vibration signal or noise. In the current industry production, in which real-time data acquisition - whether for processes control, or for health condition monitoring of equipment - is the reality, it is necessary to develop auxiliary methods that provide high reliability to identification, localization and failure prognostics. In this work, the main objective is to provide a spur gears’ model-based methodology for condition-monitoring and to validate a dynamic model with experimental vibration data of a gearbox. Hence, a dynamic model of spur meshing gears was developed considering a 6 degrees of freedom and time-varying meshing stiffness to simulate vibrations signals; a tooth profile error was also included; in this analytical model, a straight crack was simulated by reducing the meshing stiffness in a tooth; experiments with a gearbox experimental set were run; and, some signal processing was apllied in the vibration data. The results allowed the model validation with the comparison between simulate and experimental signals, in time-domain and frequency-domain / Mestre

This thesis develops an advanced battery monitoring and control system based on the electrochemical principles that govern lithium-ion battery dynamics. This work is motivated by the need of having safer and better energy storage systems for all kind of applications, from small scale portable electronics to large scale renewable energy storage. In this context, lithium-ion batteries have become the enabling technology for energy autonomy in appliances (e.g. mobile phone, electric vehicle) and energy self-consumption in households. However, batteries are oversized and pricey, might be unsafe, are slow to charge and may not equalize the lifetime of the application they are intended to power. This work tackles these different issues.This document first introduces the general context of the battery management problem, as well as the particular issues that arise when modeling, supervising and controlling the battery short-term and long-term operation. Different solutions coming from the literature are reviewed, and several standard tools borrowed from control theory are exposed. Then, starting by well-known contributions in electrochemical modeling, we proceed to develop reduced-order models for the battery operation including degradation mechanisms, that are highly descriptive of the real phenomena taking place. This modeling framework is the cornerstone of all the monitoring and control development that follows.Next, we derive a battery diagnosis system with a twofold objective. First, indicators for internal faults affecting the battery state-of-health are obtained. Secondly, detection and isolation of sensor faults is achieved. Both tasks rely on state observers designed from electrochemical models to perform state estimation and residual generation. Whereas the former solution resorts to system identification techniques for health monitoring, the latter solution exploits fault diagnosis for instrumentation assessment.We then develop a feedback battery charge strategy able to push in performance while accounting for constraints associated to battery degradation. The fast and safe charging capabilities of the proposed approach are ultimately validated through long-term cycling experiments. This approach outperforms widely used commercial charging strategies in terms of both charging speed and degradation.The main contribution of this thesis is the exploitation of first principles models to develop battery management strategies towards improving safety, charging time and lifetime of battery systems without jeopardizing performance. The obtained results show that system and control theory offer opportunities to improve battery operation, aside from the material sciences contributions to this field. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

La contribución que se pretende con esta tesis se refiere a la implantación de un sistema de diagnosis de fallos en plantas químicas completas integrado al sistema de supervisión, gestión y control de la producción.El sistema de diagnosis de fallos que se presenta consiste en una combinación de un sistema de reconocimiento de patrones basado en redes neuronales artificiales y un sistema de inferencia basado en la lógica difusa. La información necesaria para desarrollar el sistema de diagnosis incluye los datos históricos, un análisis de riesgo y operabilidad y un modelo de la planta química. La entrada al sistema son las mediciones directas o indirectas de la planta y la salida consiste en una señal para cada fallo (0: no fallo; 1: fallo). Primero se definen los fallos posibles. La red neuronal se entrena con datos históricos de fallos ocurridos en el pasado, con el objeto de reconocer los patrones respectivos. En el caso de que no se posean los datos históricos de alguno de los fallos, por ejemplo porque nunca hayan ocurrido, se obtienen los patrones mediante la simulación, usando el modelo de la planta. El sistema de lógica difusa contiene un conjunto de reglas si-entonces que pueden ser de dos tipos: las basadas en el conocimiento de la planta, mediante el análisis de riesgo o la experiencia con la simulación, y las basadas en la experiencia con el uso de la red neuronal, previamente entrenada. Otro aspecto novedoso es la posibilidad de entrenar la red neuronal con "características" extraídas de las variables medidas mediante su pre-procesamiento con wavelets. Esta variante permite obtener un alto rendimiento del sistema de diagnosis en plantas químicas discontinuas y continuas complejas. Para optimizar los parámetros de los componentes del sistema de diagnosis se propone un índice de rendimiento. Además, se utiliza el índice de rendimiento para comparar el sistema propuesto con otros métodos.Las señales dadas por el sistema de diagnosis pueden ser usadas por el sistema de programación de la producción para actualizar el plan de la manera más efectiva, por el sistema de control para actuar en forma automática y por los operadores de planta como soporte para la toma de decisiones. Se han sentado las bases para la traducción de la salida del sistema de diagnosis para su utilización por los demás niveles del soporte informático. Se usa una estrategia basada en el análisis de riesgo y operabilidad de la planta.El sistema propuesto es consecuencia de sucesivas mejoras, al trabajar con diferentes casos de estudio. Los escenarios académicos corresponden a una planta química continua con una corriente de reciclo y un reactor discontinuo. Los casos a escala de planta piloto corresponden a escenarios construidos en la UPC: un reactor de gasificación, un reactor semicontinuo y una planta discontinua multipropósito. Los escenarios industriales corresponden a dos refinerías de azúcar y a un sector de una planta petroquímica.El sistema propuesto muestra ser ventajoso respecto a otros métodos tanto en cuanto a la rapidez de diagnosis como en cuanto a su capacidad para aislar los fallos. La simplicidad del desarrollo y la flexible estrategia de implementación del sistema propuesto auguran un futuro promisorio a la tecnología presentada. Nuevas líneas de investigación se pueden emprender mediante el desarrollo de un sistema de gestión de las alarmas. Otro aspecto importante es la posibilidad de participar en la estandarización de las interfaces del programa de diagnosis. / The pretended contribution of this thesis deals with the implementation of a fault diagnosis system in chemical plants integrated to the monitoring, management and control system. The proposed fault diagnosis system consists in a combination of a pattern recognition approach based on artificial neural networks and an inference system based on fuzzy logic. The information needed to develop the fault diagnosis system includes the historical data, the hazard and operability study and the model of the chemical plant. The inputs to the system are the direct or indirect measurements from the plant and the output consists in a signal for each fault (0: no fault; 1: fault). First, the possible faults are defined. The artificial neural network is trained with historical data of faults occurred in the past, with the aim of recognising the respective patterns. In the case that the corresponding historical data are not available, for example due to the no occurrence of the fault, the patterns are obtained through simulation, using the plant model. The fuzzy logic system contains a set of if-then rules that can be of two types: those based in the process knowledge, by the hazard analysis or by the experience with simulation, and those based on the experience with the use of an artificial neural network, previously trained. Other novel aspect is the possibility of artificial neural network training by using signals features that are extracted by its pre-processing using wavelets. This alternative allows a higher fault diagnosis system performance in batch and complex continuous chemical plants. In order to optimise the parameters of the components of the fault diagnosis system, a performance index is proposed. The performance index is also utilised to compare the proposed fault diagnosis system against other methods.The signals provided by the fault diagnosis system can be used by the scheduling system to update the schedule in the most effective way, by the control system to take automated control actions and by plant's operators as support for decision-making. The basis of the translation of the system output, for its utilisation at other levels in the information system, has been settled.The proposed strategy is based on the hazard and operability analysis.The proposed system is the result of successive improvements, by working with different case studies. The academic scenarios correspond to a continuous chemical plant with a recycle stream and a batch reactor. The pilot plant scale cases correspond to scenarios built at UPC: a reactor gasifier, a fed-batch reactor and a multipurpose batch chemical plant. The industrial scenarios correspond to two sugar refineries and a sector of a petrochemical plant.The proposed system shows to be advantageous with respect to other methods in relation to the fastness of the diagnosis and also its capacity to isolate faults. The simplicity of the development and the flexible strategy of implementation of the proposed fault diagnosis system give a promising future to the presented technology. New research lines can be considered by developing the alarm handling system. Other important aspect is the possibility of the participation in the standardisation of the interfaces of the fault diagnosis program.

Power system protection is important for service reliability and quality assurance. Various faults may occur due to natural and artificial calamity. To reduce the outage duration and promptly restore power services, fault section estimate has to be done effectively and accurately with fault alarms. Dispatchers study the changed statuses of protection devices from the Supervisory Control and Data Acquisition (SCADA) system to identify the fault. Single and multiple faults could coexist with the failed operation of relays and circuit breakers, or with the erroneous data communication. It needs a long time to process a large number of alarms under various conditions involving multiple faults and many uncertainties. To cope with the problem, an effective tool is helpful for the fault section estimation and alarm processing. Besides, power transformer plays a major role in a power system. For a better service quality, it is important to be routinely examined for detecting incipient faults inside transformers. Preventive techniques for early detection can find out the incipient faults and avoid outages. Power quality is another issue to considerable attentions from utilities and customers due to the popular uses of many sensitive electronic equipment. Harmonics, voltage swell, voltage sag, and, power interruption could downgrade the service quality. To ensure the power quality, detecting harmonic and voltage disturbances becomes important. A detection method with classification capability will be helpful for detecting disturbances. This dissertation developed various algorithm for detection including fault section detection, alarm processing, transformer fault diagnosis, and power quality detection. For a well-dispatched power system, the adaptive detection idea will be used, and the existing SCADA/EMS will be integrated without extra devices.

Knowledge-Based Systems (KBSs) represent a relatively new programming approach and methodology that has evolved and is still evolving as an important sub-area of Artificial Intelligence (AI) research. The most prevalent application of KBSs, which emerged in recent times, has been various types of diagnosis and troubleshooting. KBS has an important role to play, particularly in fault diagnosis of process plants, which involve lot of challenges starting from commonly occurring malfunctions to rarely occurring emergency situations. The KBS approach is promising for this domain as it captures efficient problem-solving of experts, guides the human operator in rapid fault detection, explains the line of reasoning to the human operator, and supports modification and refinement of the process knowledge as experience is gained. However, most of the current KBSs in process plants are built on expert knowledge compiled in the form of production rules. These systems lack flexibility due to their process-specific nature and are unreliable when faced with unanticipated faults. Although attempts have been made to integrate knowledge based on experience and 'deep' process knowledge to overcome this lack of flexibility, very little work has been reported to make the diagnostic system flexible and usable for various plant configurations. In this thesis, we propose a hybrid knowledge framework which includes both process-specific and process-common knowledge of the structure and behavior of the domain, and a process-independent diagnostic mechanism based on causal and qualitative reasoning. This framework is flexible and allows a unified design methodology for fault diagnosis of process plants. The process-specific knowledge includes experiential knowledge about commonly occurring faults, behavioral knowledge about causal interactions among process-dependent variables, and structural knowledge about components' description and connectivity. The process-common knowledge comprises template models of various types of components commonly present in any process plant, constraints and confluences based on mass and energy balances between parameters across components. The process behavioral knowledge is qualitatively represented in the form of Signed Digraph (SDG), which is converted into a set of rules (SDGrules), added with control premises for the purpose of diagnostic reasoning. Frame-objects are used to represent the structural knowledge, while rules are used to capture experiential knowledge about common faults. An interface program viz., Knowledge Acquisition Interface (KAI) aids acquisition and conversion of (i) behavioral knowledge into a set of SDG-rules and (ii) structural knowledge and experience-based heuristic rules into a set of facts. The Diagnostic Mechanism is based on a steady state model of the process and is composed of three consecutive phases for locating a fault. The first phase is Malfunction Block Identification (MBT), which locates a malfunctioning subsystem or Malfunction Block (MB) that is responsible for causing the process malfunction. It is based on alarm data whenever violation of process parameters occurs. Once the suspected MB is identified, the second phase viz., Malfunction Parameter Identification (MPI) is invoked t o locate parameters which indicate the prime cause(s) of the fault in that MB. This is achieved by correlating various instrumentation data through causal relationships described by the SDG-rules of that MB. Finally, Malfunctioning Component Identification (MCI) phase is invoked to locate the malfunctioning component. MCI phase uses the malfunction parameter (s) obtained from previous phase and experiential and structural knowledge of that MA for this purpose. The Diagnostic Mechanism is process-independent and, therefore, is capable of adapting to various types of plant configurations. Since, the Knowledge Base and the Diagnostic Mechanism are separate, modification of either of them can be done independently. The Diagnostic Mechanism is potentially capable of investigating symptoms that have multiple or unrelated origins. It also provides explanation facility for justifying the line of diagnostic reasoning to the human operator.

This research develops a fault diagnosis method for complex systems in the presence of uncertainties and possibility of multiple solutions. Fault diagnosis is a challenging problem because data used in diagnosis contain random errors and often systematic errors as well. Furthermore, fault diagnosis is basically an inverse problem so that it inherits unfavorable characteristics of inverse problems: The existence and uniqueness of an inverse solution are not guaranteed and the solution may be unstable. The weighted least squares method and its variations are traditionally used for solving inverse problems. However, the existing algorithms often fail to identify multiple solutions if they are present. In addition, the existing algorithms are not capable of selecting variables systematically so that they generally use the full model in which may contain unnecessary variables as well as necessary variables. Ignoring this model uncertainty often gives rise to, so called, the smearing effect in solutions, because of which unnecessary variables are overestimated and necessary variables are underestimated. The proposed method solves the inverse problem using Bayesian inference. An engineering system can be parameterized using state variables. The probability of each state variable is inferred from observations made on the system. A bias in an observation is treated as a variable, and the probability of the bias variable is inferred as well. To take the uncertainty of model structure into account, multiple Bayesian models are created with various combinations of the state variables and the bias variables. The results from all models are averaged according to how likely each model is. Gibbs sampling is used for approximating updated probabilities. The method is demonstrated for two applications: the status matching of a turbojet engine and the fault diagnosis of an industrial gas turbine. In the status matching application only physical faults in the components of a turbojet engine are considered whereas in the fault diagnosis application sensor biases are considered as well as physical faults. The proposed method is tested in various faulty conditions using simulated measurements. Results show that the proposed method identifies physical faults and sensor biases simultaneously. It is also demonstrated that multiple solutions can be identified. Overall, there is a clear improvement in ability to identify correct solutions over the full model that contains all state and bias variables.

Today's industry uses increasingly complex machines, some with extremely demanding performance criteria. Failed machines can lead to economic loss and safety problems due to unexpected production stoppages. Fault diagnosis in the condition monitoring of these machines is crucial for increasing machinery availability and reliability. Fault diagnosis of machinery is often a difficult and daunting task. To be truly effective, the process needs to be automated to reduce the reliance on manual data interpretation. It is the aim of this research to automate this process using data from machinery vibrations. This thesis focuses on the design, development, and application of an automatic diagnosis procedure for rolling element bearing faults. Rolling element bearings are representative elements in most industrial rotating machinery. Besides, these elements can also be tested economically in the laboratory using relatively simple test rigs. Novel modern signal processing methods were applied to vibration signals collected from rolling element tests to destruction. These included three advanced timefrequency signal processing techniques, best basis Discrete Wavelet Packet Analysis (DWPA), Matching Pursuit (MP), and Basis Pursuit (BP). This research presents the first application of the Basis Pursuit to successfully diagnosing rolling element faults. Meanwhile, Best basis DWPA and Matching Pursuit were also benchmarked with the Basis Pursuit, and further extended using some novel ideas particularly on the extraction of defect related features. The DWPA was researched in two aspects: i) selecting a suitable wavelet, and ii) choosing a best basis. To choose the most appropriate wavelet function and decomposition tree of best basis in bearing fault diagnostics, several different wavelets and decomposition trees for best basis determination were applied and comparisons made. The Matching Pursuit and Basis Pursuit techniques were effected by choosing a powerful wavelet packet dictionary. These algorithms were also studied in their ability to extract precise features as well as their speed in achieving a result. The advantage and disadvantage of these techniques for feature extraction of bearing faults were further evaluated. An additional contribution of this thesis is the automation of fault diagnosis by using Artificial Neural Networks (ANNs). Most of work presented in the current literature has been concerned with the use of a standard pre-processing technique - the spectrum. This research employed additional pre-processing techniques such as the spectrogram and DWPA based Kurtosis, as well as the MP and BP features that were subsequently incorporated into ANN classifiers. Discrete Wavelet Packets and Spectra, were derived to extract features by calculating RMS (root mean square), Crest Factor, Variance, Skewness, Kurtosis, and Matched Filter. Certain spikes in Matching Pursuit analysis and Basis Pursuit analysis were also used as features. These various alternative methods of pre-processing for feature extraction were tested, and evaluated with the criteria of the classification performance of Neural Networks. Numerous experimental tests were conducted to simulate the real world environment. The data were obtained from a variety of bearings with a series of fault severities. The mechanism of bearing fault development was analysed and further modelled to evaluate the performance of this research methodology. The results of the researched methodology are presented, discussed, and evaluated in the results and discussion chapter of this thesis. The Basis Pursuit technique proved to be effective in diagnostic tasks. The applied Neural Network classifiers were designed as multi layer Feed Forward Neural Networks. Using these Neural Networks, automatic diagnosis methods based on spectrum analysis, DWPA, Matching Pursuit, and Basis Pursuit proved to be effective in diagnosing different conditions such as normal bearings, bearings with inner race and outer race faults, and rolling element faults, with high accuracy. Future research topics are proposed in the final chapter of the thesis to provide perspectives and suggestions for advancing research into fault diagnosis and condition monitoring.

Made available in DSpace on 2014-06-11T19:30:50Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-07-21Bitstream added on 2014-06-13T19:19:31Z : No. of bitstreams: 1 leao_fb_dr_ilha.pdf: 4326970 bytes, checksum: 5e80d8b3eb8a0bff2c52ea28e2f0a451 (MD5) / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Neste trabalho é proposta uma metodologia para a análise e interpretação de alarmes em tempo real em sistemas de distribuição de energia elétrica, considerando o diagnóstico em nível de subestações e redes. A metodologia busca superar as dificuldades e desvantagens dos métodos já propostos na literatura especializada para resolver o diagnóstico de faltas em sistemas de potência. O método proposto emprega um modelo matemático original bem como um novo algoritmo genético para efetuar o diagnóstico dos alarmes de maneira eficiente e rápida. O modelo matemático é dividido em duas partes fundamentais: (1) modelo de operação do sistema de proteção; e (2) modelo de Programação Binária Irrestrita (PBI). A parte (1) é composta por um conjunto de equações de estados esperados das funções de proteção dos relés do sistema, modeladas com base na lógica de operação de funções de proteção tais como sobrecorrente, diferencial e distância, bem como na filosofia de proteção de sistemas de potência. A parte (2) é estabelecida através de uma função objetivo formulada com base na teoria de cobertura parcimoniosa (parcimonious set covering theory), e busca a associação ou “match” entre os relatórios de alarmes informados pelo sistema SCADA (Supervisory Control and Data Acquisition) e os estados esperados das funções de proteção formuladas na parte (1) do modelo. O novo algoritmo genético proposto é empregado para minimizar o modelo de PBI e possui como característica a utilização de dois parâmetros de controle. O algoritmo possui taxas de recombinação e mutação automática e dinamicamente calibradas, baseadas na saturação da população corrente, possuindo uma imediata resposta à possível convergência prematura para ótimos locais. A metodologia desenvolvida para o diagnóstico... / This work proposes a methodology for the analysis and interpretation of real-time alarms in electric power distribution systems in the substation level and network level. The methodology seeks to overcome the difficulties and disadvantages of the methods already proposed in the literature to solve the fault diagnosis in power systems. The proposed method employs a novel mathematical model and a genetic algorithm to carry out the diagnosis of alarms efficiently and quickly. The model is divided into two main parts: (1) a protection system operation model; and (2) Unconstrained Binary Programming (UBP) model. Part (1) provides a set of expected state equations of the protective relay functions established based on the protection operation logic such as overcurrent, differential and distance as well as the protection philosophy. Part (2) is established through an objective function formulated based on parsimonious set covering theory for associating the alarms reported by SCADA (Supervisory Control and Data Acquisition) system with the expected states of the protective relay functions. The novel genetic algorithm use only two control parameters and is employed to minimize the UBP model. In addition the algorithm has recombination and mutation rates automatically and dynamically calibrated based on the saturation of the current population and it presents an immediate response to possible premature convergence to local optima. The methodology developed for the diagnosis of substations is extended to distribution networks considering that the network has sufficient level of automation for remote monitoring of the primary feeders. In this way a new paradigm for protection of distribution networks developed based on Smart Grid concept is proposed. Extensive tests are performed with the methodology applied to distribution... (Complete abstract click electronic access below)

Made available in DSpace on 2016-12-12T17:38:33Z (GMT). No. of bitstreams: 1 Daniel Gumiero Noronha Maas.pdf: 6258702 bytes, checksum: 06207c47266ccded357fbb819c6ae9bf (MD5) Previous issue date: 2014-09-09 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / This work presents a multilayer architecture for fault diagnosis in embedded systems that allows a better discrimination of the type and source of the failure, providing an accurate and assertive diagnosis. This architecture contemplates the necessary interfaces to allow integration of this diagnostic framework in the embedded system and also considers the treatment of diagnostic information for recovery system actions purposes, or simply allows the externalization of such information. This work considers the diagnosers designed according to the methodology of fault diagnosis in DES modeled by automata. Once designed the diagnosers are implemented in ANSI C language through an automated software generation tool, and finally incorporated into the main product software where it intends to perform the diagnosis. This architecture diagnosis was then applied in a case study for Frost Free refrigerator, for which the diagnosers were designed then were implemented in software and subsequently validated in order to confirm the effectiveness of the diagnosers, of the proposed architecture beyond the C language conversion process. / Este trabalho apresenta uma arquitetura de diagnóstico multicamadas para detecção de falhas em sistemas embarcados, que permite uma melhor discriminação do tipo e origem da falha, possibilitando um diagnóstico mais preciso e assertivo. Esta arquitetura contempla as interfaces necessárias para permitir a integração no sistema embarcado e também considera o tratamento das informações de diagnóstico para fins de ações de recuperação do sistema, ou simplesmente a externalização destas informações. Neste trabalho, consideram-se os diagnosticadores projetados conforme a metodologia de diagnóstico de falhas em SEDs modelados por autômatos. Uma vez concebidos, os diagnosticadores são implementados em linguagem ANSI C, através de uma ferramenta de geração automática de software, e finalmente incorporados ao software principal do equipamento onde se pretende realizar o diagnóstico. Esta arquitetura de diagnóstico foi então aplicada em um estudo de caso para um refrigerador Frost Free, para o qual foram projetados os diagnosticadores, em seguida os mesmos foram implementados em software e posteriormente validados a fim de comprovar a eficácia não somente dos diagnosticadores mas também da arquitetura proposta, além do processo de conversão dos mesmos em linguagem de software.