Fault detection and diagnosis is an important problem in process engineering. In this dissertation, use of multivariate techniques for fault detection and diagnosis is explored in the context of statistical process control. Principal component analysis and its extension, kernel principal component analysis, are proposed to extract features from process data. Kernel based methods have the ability to model nonlinear processes by forming higher dimensional representations of the data. Discriminant methods can be used to extend on feature extraction methods by increasing the isolation between different faults. This is shown to aid fault diagnosis. Linear and kernel discriminant analysis are proposed as fault diagnosis methods. Data from a pilot scale distillation column were used to explore the performance of the techniques. The models were trained with normal and faulty operating data. The models were tested with unseen and/or novel fault data. All the techniques demonstrated at least some fault detection and diagnosis ability. Linear PCA was particularly successful. This was mainly due to the ease of the training and the ability to relate the scores back to the input data. The attributes of these multivariate statistical techniques were compared to the goals of statistical process control and the desirable attributes of fault detection and diagnosis systems. / Dissertation (MEng (Control Engineering))--University of Pretoria, 2008. / Chemical Engineering / MEng / Unrestricted

Modern electric machine drives, particularly three phase permanent magnet machine drive systems represent an indispensable part of high power density products. Such products include; hybrid electric vehicles, large propulsion systems, and automation products. Reliability and cost of these products are directly related to the reliability and cost of these systems. The compatibility of the electric machine and its drive system for optimal cost and operation has been a large challenge in industrial applications. The main objective of this dissertation is to find a design and control scheme for the best compromise between the reliability and optimality of the electric machine-drive system. The effort presented here is motivated by the need to find new techniques to connect the design and control of electric machines and drive systems. A highly accurate and computationally efficient modeling process was developed to monitor the magnetic, thermal, and electrical aspects of the electric machine in its operational environments. The modeling process was also utilized in the design process in form finite element based optimization process. It was also used in hardware in the loop finite element based optimization process. The modeling process was later employed in the design of a very accurate and highly efficient physics-based customized observers that are required for the fault diagnosis as well the sensorless rotor position estimation. Two test setups with different ratings and topologies were numerically and experimentally tested to verify the effectiveness of the proposed techniques. The modeling process was also employed in the real-time demagnetization control of the machine. Various real-time scenarios were successfully verified. It was shown that this process gives the potential to optimally redefine the assumptions in sizing the permanent magnets of the machine and DC bus voltage of the drive for the worst operating conditions. The mathematical development and stability criteria of the physics-based modeling of the machine, design optimization, and the physics-based fault diagnosis and the physics-based sensorless technique are described in detail. To investigate the performance of the developed design test-bed, software and hardware setups were constructed first. Several topologies of the permanent magnet machine were optimized inside the optimization test-bed. To investigate the performance of the developed sensorless control, a test-bed including a 0.25 (kW) surface mounted permanent magnet synchronous machine example was created. The verification of the proposed technique in a range from medium to very low speed, effectively show the intelligent design capability of the proposed system. Additionally, to investigate the performance of the developed fault diagnosis system, a test-bed including a 0.8 (kW) surface mounted permanent magnet synchronous machine example with trapezoidal back electromotive force was created. The results verify the use of the proposed technique under dynamic eccentricity, DC bus voltage variations, and harmonic loading condition make the system an ideal case for propulsion systems.

O escopo desta dissertação é o desenvolvimento de uma metodologia para a detecção e diagnóstico de falhas em processos industriais baseado em modelos empíricos no subespaço das variáveis do processo com expansão não linear das bases. A detecção e o diagnóstico de falhas são fundamentais para aumentar a segurança, confiabilidade e lucratividade de processos industriais. Métodos qualitativos, quantitativos e baseados em dados históricos do processo têm sido estudados amplamente. Para demonstrar as vantagens da metodologia proposta, ela será comparada com duas metodologias consideradas padrão, uma baseada em Análise de Componentes Principais (PCA) e a outra baseada em Mínimos Quadrados Parciais (PLS). Dois estudos de casos são empregados nessa comparação. O primeiro consiste em um tanque de aquecimento com mistura e o segundo contempla o estudo de caso do processo da Tennessee Eastman. As vantagens da metodologia proposta consistem na redução da dimensionalidade dos dados a serem usados para um diagnóstico adequado, além de detectar efetivamente a anormalidade e identificar as variáveis mais relacionadas à falha, permitindo um melhor diagnóstico. Além disso, devido à expansão das bases dos modelos é possível trabalhar efetivamente com sistemas não lineares, através de funções polinomiais e exponenciais dentro do modelo. Adicionalmente o trabalho contém uma metodologia de validação dos resultados da metodologia proposta, que consiste na eliminação das variáveis do melhor modelo obtido pelos Modelos Empíricos, através do método Backward Elimination. A metodologia proposta forneceu bons resultados na área do diagnóstico de falhas: conseguiu-se uma grande diminuição da dimensionalidade nos sistemas estudados em até 93,55%, bem como uma correta detecção de anormalidades e permitiu a determinação das variáveis mais relacionadas às anormalidades do processo. As comparações feitas com as metodologias padrões permitiram demonstrar que a metodologia proposta tem resultados superiores, pois consegue detectar as anormalidades em um espaço dimensional reduzido, detectando comportamentos não lineares e diminuindo incertezas. / Fault detection and diagnosis are critical to increasing the safety, reliability, and profitability of industrial processes. Qualitative and quantitative methods and process historical data have been extensively studied. This article proposes a methodology for fault detection and diagnosis, based on historical data of processes and the creation of empirical models with the expansion of nonlinear bases (polynomial and exponential bases) and regularization techniques. To demonstrate the advantages of the proposed approach, it is compared with two standard methodologies: Principal Components Analysis (PCA) and the Partial Least Squares (PLS), performed in two case studies: a mixed heating tank and the Tennessee Eastman Process. The advantages of the proposed methodology are the reduction of the dimensionality of the data used, in addition to the effective detection of abnormalities, identifying the variables most related to the fault. Furthermore, the work contains a methodology to validate the diagnosis results consisting of variable elimination from the best empirical models with the Backward Elimination algorithm. The proposed methodology achieved a promising performance, since it can decrease the dimensionality of the studied systems up to 93.55%, reducing uncertainties, and capturing nonlinear behaviors.

The substantial growth in the use of automated in-process sensing technologies creates great opportunities for manufacturers to detect abnormal manufacturing processes and identify the root causes quickly. It is critical to locate and distinguish two types of faults - process faults and sensor faults. The procedures to monitor and diagnose process and sensor mean shift faults are presented with the assumption that the manufacturing processes can be modeled by a linear fault-quality model. A W control chart is developed to monitor the manufacturing process and quickly detect the occurrence of the sensor faults. Since the W chart is insensitive to process faults, when it is combined with U chart, both process faults and sensor faults can be detected and distinguished. A unit-free index referred to as the sensitivity ratio (SR) is defined to measure the sensitivity of the W chart. It shows that the sensitivity of the W chart is affected by the potential influence of the sensor measurement. A Bayesian variable selection based fault diagnosis approach is presented to locate the root causes of the abnormal processes. A Minimal Coupled Pattern (MCP) and its degree are defined to denote the coupled structure of a system. When less than half of the faults within an MCP occur, which is defined as sparse faults, the proposed fault diagnosis procedure can identify the correct root causes with high probability. Guidelines are provided for the hyperparameters selection in the Bayesian hierarchical model. An alternative CML method for hyperparameters selection is also discussed. With the large number of potential process faults and sensor faults, an MCMC method, e.g. Metropolis-Hastings algorithm can be applied to approximate the posterior probabilities of candidate models. The monitor and diagnosis procedures are demonstrated and evaluate through an autobody assembly example.

In a modern diesel engine, a fuel injector is a vital component. The injectors control the fuel dosing into the combustion chambers. The accuracy in the fuel dosing is very important as inaccuracies have negative effects on engine out emissions and the controllability. Because of this, a diagnosis that can classify the conditions of the injectors with good accuracy is highly desired. A signal that contains information about the injectors condition, is the engine angular velocity. In this thesis, the classification performance of six common machine learning methods is evaluated. The input to the methods is the engine angular velocity. In addition to the classification performance, also the computational cost of the methods, in a deployed state, is analysed. The methods are evaluated on data from a Scania truck that has been run just like any similar commercial vehicle. The six methods evaluated are: logistic regression, kernel logistic regression, linear discriminant analysis, quadratic discriminant analysis, fully connected neural networks and, convolutional neural networks. The results show that the neural networks achieve the best classification performance. Furthermore, the neural networks also achieve the best classification performance from a, in a deployed state, computational cost effectiveness perspective. Results also indicate that the neural networks can avoid false alarms and maintain high sensitivity.

As the automotive industry constantly makes technological progress, higher demands are placed on safety, environmentally friendly and durability. Modern vehicles are headed towards increasingly complex system, in terms of both hardware and software making it important to detect faults in any of the components. Monitoring the engine’s health has traditionally been done using expert knowledge and model-based techniques, where derived models of the system’s nominal state are used to detect any deviations. However, due to increased complexity of the system this approach faces limitations regarding time and knowledge to describe the engine’s states. An alternative approach is therefore data-driven methods which instead are based on historical data measured from different operating points that are used to draw conclusion about engine’s present state. In this thesis a proposed diagnostic framework is presented, consisting of a systematically approach for fault classification of known and unknown faults along with a fault size estimation. The basis for this lies in using principal component analysis to find the fault vector for each fault class and decouple one fault at the time, thus creating different subspaces. Importantly, this work investigates the efficiency of taking multiple classifiers into account in the decision making from a performance perspective. Aggregating multiple classifiers is done solving a quadratic optimization problem. To evaluate the performance, a comparison with a random forest classifier has been made. Evaluation with challenging test data show promising results where the algorithm relates well to the performance of random forest classifier.

Accurate and timely diagnosis of faults is essential for the reliability and security of power grid operation and maintenance. The emergence of big data has enabled the incorporation of a vast amount of information in order to create custom fault datasets and improve the diagnostic capabilities of existing frameworks. Intelligent systems have been successful in incorporating big data to improve diagnostic performance using computational intelligence and machine learning based on fault datasets. Among these systems are fuzzy inference systems with the ability to tackle the ambiguities and uncertainties of a variety of input data such as climate data. This makes these systems a good choice for extracting knowledge from energy big data. In this thesis, qualitative climate information is used to construct a fault dataset. A fuzzy inference system is designed whose parameters are optimized using a single layer artificial neural network. This fault diagnosis framework maps the relationship between fault variables in the fault dataset and fault types in real-time to improve the accuracy and cost efficiency of the framework. / Graduate

L'objectif principal de cette thèse est de développer un algorithme dynamique de partitionnement de données (classification non supervisée ou " clustering " en anglais) qui ne se limite pas à des concepts statiques et qui peut gérer des distributions qui évoluent au fil du temps. Cet algorithme peut être utilisé dans les systèmes de surveillance du processus, mais son application ne se limite pas à ceux-ci. Les contributions de cette thèse peuvent être présentées en trois groupes: 1. Contributions au partitionnement dynamique de données en utilisant : un algorithme de partitionnement dynamique basé à la fois sur la distance et la densité des échantillons est présenté. Cet algorithme ne fait aucune hypothèse sur la linéarité ni la convexité des groupes qu'il analyse. Ces clusters, qui peuvent avoir des densités différentes, peuvent également se chevaucher. L'algorithme développé fonctionne en ligne et fusionne les étapes d'apprentissage et de reconnaissance, ce qui permet de détecter et de caractériser de nouveaux comportements en continu tout en reconnaissant l'état courant du système. 2. Contributions à l'extraction de caractéristiques : une nouvelle approche permettant d'extraire des caractéristiques dynamiques est présentée. Cette approche, basée sur une approximation polynomiale par morceaux, permet de représenter des comportements dynamiques sans perdre les informations relatives à la magnitude et en réduisant simultanément la sensibilité de l'algorithme au bruit dans les signaux analysés. 3. Contributions à la modélisation de systèmes à événements discrets évolutifs a partir des résultats du clustering : les résultats de l'algorithme de partitionnement sont utilisés comme base pour l'élaboration d'un modèle à événements discrets du processus. Ce modèle adaptatif offre une représentation du comportement du processus de haut niveau sous la forme d'un automate dont les états représentent les états du processus appris par le partitionnement jusqu'à l'instant courant et les transitions expriment l'atteignabilité des états. / The main objective of this thesis is to propose a dynamic clustering algorithm that can handle not only dynamic data but also evolving distributions. This algorithm is particularly fitted for the monitoring of processes generating massive data streams, but its application is not limited to this domain. The main contributions of this thesis are: 1. Contribution to dynamic clustering by the proposal of an approach that uses distance- and density-based analyses to cluster non-linear, non-convex, overlapped data distributions with varied densities. This algorithm, that works in an online fashion, fusions the learning and lassification stages allowing to continuously detect and characterize new concepts and at the same time classifying the input samples, i.e. which means recognizing the current state of the system in a supervision application. 2. Contribution to feature extraction by the proposal of a novel approach to extract dynamic features. This approach ,based on piece-polynomial approximation, allows to represent dynamic behaviors without losing magnitude related information and to reduce at the same time the algorithm sensitivity to noise corrupting the signals. 3. Contribution to automatic discrete event modeling for evolving systems by exploiting informations brought by the clustering. The generated model is presented as a timed automaton that provides a high-level representation of the behavior of the process. The latter is adaptive in the sense that its construction is elaborated following the discovery of new concepts by the clustering algorithm.

No description available.

No description available.