No description available.

No description available.

The new approach for sensor diagnostics is presented. The approach, Enhanced Autoassociative Neural Networks (E-AANN), adds enhancement to Autoassociative Neural Networks (AANN) developed by Kramer in 1992. This enhancement allows AANN to identify faulty sensors. E-AANN uses a secondary optimization process to identify and reconstruct sensor faults. Two common types of sensor faults are investigated, drift error and shift or offset error. In the case of drift error, the sensor error occurs gradually while in the case of shift error, the sensor error occurs abruptly. EAANN catches these error types. A chiller model provided synthetic data to test the diagnostic approach under various noise level conditions. The results show that sensor faults can be detected and corrected in noisy situations with the E-AANN method described. In high noisy situations (10% to 20% noise level), E-AANN performance degrades. E-AANN performance in simple dynamic systems was also investigated. The results show that in simple dynamic situations, E-AANN identifies faulty sensors.

Process monitoring in the chemical and other process industries has been of great practical importance. Early detection of faults is critical in avoiding product quality deterioration, equipment damage, and personal injury. The goal of this dissertation is to develop process monitoring schemes that can be applied to complex process systems. Neural networks have been a popular tool for modeling and pattern classification for monitoring of process systems. However, due to the prohibitive computational cost caused by high dimensionality and frequently changing operating conditions in batch processes, their applications have been difficult. The first part of this work tackles this problem by employing a polynomial-based data preprocessing step that greatly reduces the dimensionality of the neural network process model. The process measurements and manipulated variables go through a polynomial regression step and the polynomial coefficients, which are usually of far lower dimensionality than the original data, are used to build a neural network model to produce residuals for fault classification. Case studies show a significant reduction in neural model construction time and sometimes better classification results as well. The second part of this research investigates classification trees as a promising approach to fault detection and classification. It is found that the underlying principles of classification trees often result in complicated trees even for rather simple problems, and construction time can excessive for high dimensional problems. Fisher Discriminant Analysis (FDA), which features an optimal linear discrimination between different faults and projects original data on to perpendicular scores, is used as a dimensionality reduction tool. Classification trees use the scores to separate observations into different fault classes. A procedure identifies the order of FDA scores that results in a minimum tree cost as the optimal order. Comparisons to other popular multivariate statistical analysis based methods indicate that the new scheme exhibits better performance on a benchmarking problem.

<p>Traditional diagnosis has been performed with hardware redundancy and limit checking. The development of more powerful computers have made a new kind of diagnosis possible. Todays computing power allows models of the system to be run in real time and thus making model-based diagnosis possible. </p><p>The objective with this thesis is to investigate the potential of model-based diagnosis, especially when combined with active diagnosis. The diagnosis system has been applied on a model of the JAS39 Gripen fuel pressurization system. </p><p>With the sensors available today no satisfying diagnosis system can be built, however, by adding a couple of sensors and using active model-based diagnosis all faults can be detected and isolated into a group of at most three components. </p><p>Since the diagnosis system in this thesis only had a model of the real system to be tested at, this thesis is not directly applicable on the real system. What can be used is the diagnosis approach and the residuals and decision structure developed here.</p>

Siemens gas turbines are monitored and controlled by a large number of sensors and actuators. Process information is stored in a database and used for offline calculations and analyses. Before storing the sensor readings, a compression algorithm checks the signal and skips the values that explain no significant change. Compression of 90 % is not unusual. Since data from the database is used for analyses and decisions are made upon results from these analyses it is important to have a system for validating the data in the database. Decisions made on false information can result in large economic losses. When this project was initiated no sensor validation system was available. In this thesis the uncertainties in measurement chains are revealed. Methods for fault detection are investigated and finally the most promising methods are put to the test. Linear relationships between redundant sensors are derived and the residuals form an influence structure allowing the faulty sensor to be isolated. Where redundant sensors are not available, a gas turbine model is utilized to state the input-output relationships so that estimates of the sensor outputs can be formed. Linear parametric models and an ANN (Artificial Neural Network) are developed to produce the estimates. Two techniques for the linear parametric models are evaluated; prediction and simulation. The residuals are also evaluated in two ways; direct evaluation against a threshold and evaluation with the CUSUM (CUmulative SUM) algorithm. The results show that sensor validation using compressed data is feasible. Faults as small as 1% of the measuring range can be detected in many cases.

Ensuring the reliable operation of engineering systems has long been a subject of great practical and academic interest. This interest is clearly demonstrated by the preponderance of literature in the area of Fault Detection and Diagnosis (FDD) and Fault Tolerant Control (FTC), spanning the past three decades. However, increasingly stringent performance and safety requirements have led to engineering systems with progressively increasing complexity. This complexity has rendered many traditional FDD and FTC methods exceedingly cumbersome, often to the point of infeasibility. This thesis aims to enable FDD and FTC for complex engineering systems of interacting dynamic subsystems. For such systems, generic FDD/FTC methods have remained elusive. Effects caused by nonlinearities, interactions between subsystems and varying usage patterns complicate FDD and FTC. The goal of this thesis is to develop methods for FDD and FTC that will allow decoupling of anomalies occurred inside the monitored system from those occurred in the systems affecting the monitored system, as well as enabling performance recovery of the monitored system. In pursuit of these goals, FDD and FTC methods are explored that can account for operating regime-dependent effects in monitoring, diagnosis, prognosis and performance recovery for two classes of machines: those that operate in modes that can change only at distinct times (which often occur in manufacturing opera- tions such as drilling, milling, turning) and for those that operate in regimes that are continuously varying (such as automotive systems or electric motors). For machines that operate in modes that can change only at distinct times, a degradation model is postulated which describes how the system degrades over time for each operating regime. Using the framework of Hidden Markov Models (HMMs), modeling and identification tools are developed that enable identification a HMM of degradation for each machine operation. In the sequel, monitoring and prognosis methods that naturally follow from the framework of HMMs are also presented. The modeling and monitoring methodology is then applied to a real-world semiconductor manufacturing process using data provided by a major manufacturer. For machines that operate in regimes that are continuously varying, a behavioral model is postulated that describes the input-output dynamics of the nor- mal system in different operating regimes. Monitoring methods are presented that have the capability to account for operating regime-dependent modeling accuracies and isolate faults that have not been anticipated and for which no fault models are available. By conducting fault detection in a regime-dependent fashion, changes in modeling errors that are due to operating regime changes can be successfully distinguished from changes that are due to truly faulty operation caused by changes in the system dynamics. Enabled by this, unanticipated faults can be isolated through proliferation of the fault detection through the various subsystems of the anoma- lous system. The FDD methodology is applied to detect and diagnose faults for a multiple-input multiple-output Exhaust Gas Recirculation system in a diesel engine. Finally, methods to facilitate the recovery of normal system behavior are detailed. Using the same local model structure that was pursued for behavioral models, it is envisioned that the nominal controller will be reconfigured to attempt to recover nominal behavior as much as possible. To enable this reconfiguration, methods for automated design of closed-loop controllers for the local modeling structure are presented. Using a model-predictive approach with rigorous stability considerations, it is shown that the controllers can track a reference trajectory. Such a trajectory could be generated by any model that satisfies the control objectives, for normal or faulty systems. The controllers are then demonstrated on a benchmark nonlinear system that is nonlinear in the control. / text

Electromechanical Actuators (EMAs) are being increasingly used in many applications. There is a need to augment good design of EMAs with continuous awareness of their operational capability and make them ‘intelligent’ for two key objectives: enhancing performance to address exigent task requirements and to track any changes from their ‘as-built and certified’ state for condition-based maintenance. These objectives are achieved using a decision making philosophy where the human system operator supervises EMA operation using performance criteria and decision surfaces; updated by in-situ measurement of the variables of interest via a suite of diverse sensors. However, operational decisions made on the basis of faulty data could result in unwelcome consequences. With unexpected variations in a sensor’s output from its anticipated values, the challenge is to determine if it indicates a problem in the sensor or the monitored system. In addressing this conundrum, it is also essential to account for the inherent uncertainties present in the values being analyzed. To this end, this dissertation presents the development of a novel Sensor and Process Fault Detection and Isolation (SPFDI) algorithm. This provides a framework to utilize data from all the available sensors in a holistic manner to detect any faults in individual sensors or the system components concurrently. The algorithm uses a Bayesian network to model a system; populated with extensive empirical data. The probabilistic foundations of this method allow for incorporating and propagating uncertainties. The construction of a modular testbed and its Bayesian network are discussed in detail. Several design/ operational criteria have been proposed to aid in the creation of more usable networks in the future. The SPFDI algorithm estimates multiple values for each measurand using different combinations of input variables and probabilistic inferencing. These values are compared against those indicated by the corresponding sensors; a difference between them is indicative of a potential problem. Quantitative indicators to track the condition of different system components and sensors, termed as belief values, are modified after each comparison. The final belief values obtained at the end of an iteration of the algorithm provide a definitive indication of the sources of anomalies in the observed data and can provide guidance to the operator on decisions such as whether or not to use data from a particular sensor for updating existing decision surfaces. The representative examples and experimental results confirm the efficacy of the algorithm in detecting and isolating single as well as multiple sensor faults. The algorithm has also been found to be capable of distinguishing between sensor and system/process faults. Special categories of faults and factors that influence the execution characteristics and quality of results from the algorithm were also explored meticulously and suitable modifications have been suggested to enable the algorithm to continue to function effectively in these situations. To demonstrate the flexibility of the proposed SPFDI algorithm, its potential utilization in four broad classes of applications consisting of complex systems monitored by multiple sensors was also explored in this report. / text

Fault management is a continuously demanded function in any kind of network management. Commonly it is carried out by a centralized entity on the network which correlates collected information into likely diagnoses of the current system states. We survey the use of active-on-demand-measurement, often called active probes, together with passive readings from the perspective of one single node. The solution is confined to the node and is isolated from the surrounding environment. The utility for this approach, to fault diagnosis, was found to depend on the environment in which the specific node was located within. Conclusively, the less environment knowledge, the more useful this solution presents. Consequently this approach to fault diagnosis offers limited opportunities in the test environment. However, greater prospects was found for this approach while located in a heterogeneous customer environment.

Στην παρούσα διπλωματική εργασία ασχοληθήκαμε με το θέμα της μελέτης και διάγνωσης σφαλμάτων λόγω δυναμικής εκκεντρότητας σε Ασύγχρονη Μηχανή. Μοντελοποιήθηκε και αναλύθηκε η λειτουργία ενός τριφασικού ασύγχρονου τετραπολικού κινητήρα κλωβού, ισχύος 4 kW με 36 αυλακώσεις στο στάτη και 28 αυλακώσεις στο δρομέα, τόσο σε υγιή κατάσταση όσο και υπό συνθήκες σφάλματος. Πιο συγκεκριμένα, με χρήση της μεθόδου των Πεπερασμένων Στοιχείων σε δύο διαστάσεις μέσω του προγράμματος Opera, προσομοιώθηκε η λειτουργία του κινητήρα μας υπό δύο διαφορετικά ποσοστά σφάλματος, για 20% και για 40% και εφαρμόστηκε η μέθοδος της “Motor Current Signature Analysis - MCSA” για τη διάγνωση του σφάλματος αυτού υπό διαφορετικά επίπεδα φορτίου. / In this paper we addressed the issue of diagnostic research due to dynamic eccentricity related faults in Induction Machines. We modeled and analyzed the function of a three-phase quadrupole induction motor with a cage rotor with 36 stator slots and 28 rotor slots, both in healthy condition and under fault conditions. More specifically, using the finite element method in two dimensions through the "Opera" software, we simulated our motor operation under our two different error rates for 20% and 40% and subsequently we used the "Motor Current Signature Analysis - MCSA" diagnostic method for the diagnosis of this error under different load levels.