Browsing by Subject "Fault Detection"
Now showing 1 - 6 of 6
Results Per Page
Sort Options
Item A Comparison of Fault Detection Methods For a Transcritical Refrigeration System(2012-10-19) Janecke, Alex KarlWhen released into the atmosphere, traditional refrigerants contribute to climate change several orders of magnitude more than a corresponding amount of carbon dioxide. For that reason, an increasing amount of interest has been paid to transcritical vapor compression systems in recent years, which use carbon dioxide as a refrigerant. Vapor compression systems also impact the environment through their consumption of energy. This can be greatly increased by faulty operation. Automated techniques for detecting and diagnosing faults have been widely tested for subcritical systems, but have not been applied to transcritical systems. These methods can involve either dynamic analysis of the vapor compression cycle or a variety of algorithms based on steady state behavior. In this thesis, the viability of dynamic fault detection is tested in relation to that of static fault detection for a transcritical refrigeration system. Step tests are used to determine that transient behavior does not give additional useful information. The same tests are performed on a subcritical air-conditioner showing little value in dynamic fault detection. A static component based method of fault detection which has been applied to subcritical systems is also tested for all pairings of four faults: over/undercharge, evaporator fouling, gas cooler fouling, and compressor valve leakage. This technique allows for low cost measurement and independent detection of individual faults even when multiple faults are present. Results of this method are promising and allow distinction between faulty and fault-free behavior.Item Analysis of incipient fault signatures in inductive loads energized by a common voltage bus(Texas A&M University, 2006-04-12) Bade, Rajesh KumarRecent research has demonstrated the use of electrical signature analysis (ESA), that is, the use of induction motor currents and voltages, for early detection of motor faults in the form of embedded algorithms. In the event of multiple motors energized by a common voltage bus, the cost of installing and maintaining fault monitoring and detection devices on each motor may be avoided, by using bus level aggregate electrical measurements to assess the health of the entire population of motors. In this research an approach for detecting commonly encountered induction motor mechanical faults from bus level aggregate electrical measurements is investigated. A mechanical fault indicator is computed processing the raw electrical measurements through a series of signal processing algorithms. Inference of an incipient fault is made by the percentage relative change of the fault indicator from the ??healthy?? baseline, thus defining a Fault Indicator Change (FIC). To investigate the posed research problem, healthy and faulty motors with broken rotor bar faults are simulated using a detailed transient motor model. The FIC based on aggregate electrical measurements is studied through simulations of different motor banks containing the same faulty motor. The degradation in the FIC when using aggregate measurements, as compared to using individual motor measurements, is investigated. For a given motor bank configuration, the variation in FIC with increasing number of faulty motors is also studied. In addition to simulation studies experimental results from a two-motor setup are analyzed. The FIC and degradation in the FIC in the case of load eccentricity fault, and a combination of shaft looseness and bearing damage is studied through staged fault experiments in the laboratory setup. In this research, the viability of using bus level aggregate electrical measurements for detecting incipient faults in motors energized by a common voltage bus is demonstrated. The proposed approach is limited in that as the power rating fraction of faulty motors to healthy motors in a given configuration decreases, it becomes far more difficult to detect the presence of incipient faults at very early stages.Item Efficient Detection on Stochastic Faults in PLC Based Automated Assembly Systems With Novel Sensor Deployment and Diagnoser Design(2012-07-16) Wu, ZhenhuaIn this dissertation, we proposed solutions on novel sensor deployment and diagnoser design to efficiently detect stochastic faults in PLC based automated systems First, a fuzzy quantitative graph based sensor deployment was called upon to model cause-effect relationship between faults and sensors. Analytic hierarchy process (AHP) was used to aggregate the heterogeneous properties between sensors and faults into single edge values in fuzzy graph, thus quantitatively determining the fault detectability. An appropriate multiple objective model was set up to minimize fault unobservability and cost while achieving required detectability performance. Lexicographical mixed integer linear programming and greedy search were respectively used to optimize the model, thus assigning the sensors to faults. Second, a diagnoser based on real time fuzzy Petri net (RTFPN) was proposed to detect faults in discrete manufacturing systems. It used the real time PN to model the manufacturing plant while using fuzzy PN to isolate the faults. It has the capability of handling uncertainties and including industry knowledge to diagnose faults. The proposed approach was implemented using Visual Basic, and tested as well as validated on a dual robot arm. Finally, the proposed sensor deployment approach and diagnoser were comprehensively evaluated based on design of experiment techniques. Two-stage statistical analysis including analysis of variance (ANOVA) and least significance difference (LSD) were conducted to evaluate the diagnosis performance including positive detection rate, false alarm, accuracy and detect delay. It illustrated the proposed approaches have better performance on those evaluation metrics. The major contributions of this research include the following aspects: (1) a novel fuzzy quantitative graph based sensor deployment approach handling sensor heterogeneity, and optimizing multiple objectives based on lexicographical integer linear programming and greedy algorithm, respectively. A case study on a five tank system showed that system detectability was improved from the approach of signed directed graph's 0.62 to the proposed approach's 0.70. The other case study on a dual robot arm also show improvement on system's detectability improved from the approach of signed directed graph's 0.61 to the proposed approach's 0.65. (2) A novel real time fuzzy Petri net diagnoser was used to remedy nonsynchronization and integrate useful but incomplete knowledge for diagnosis purpose. The third case study on a dual robot arm shows that the diagnoser can achieve a high detection accuracy of 93% and maximum detection delay of eight steps. (3) The comprehensive evaluation approach can be referenced by other diagnosis systems' design, optimization and evaluation.Item Fault Detection in Dynamic Systems Using the Largest Lyapunov Exponent(2012-10-19) Sun, YifuA complete method for calculating the largest Lyapunov exponent is developed in this thesis. For phase space reconstruction, a time delay estimator based on the average mutual information is discussed first. Then, embedding dimension is evaluated according to the False Nearest Neighbors algorithm. To obtain the parameters of all of the sub-functions and their derivatives, a multilayer feedforward neural network is applied to the time series data, after the time delay and embedding dimension are fixed. The Lyapunov exponents can be estimated using the Jacobian matrix and the QR decomposition. The possible applications of this method are then explored for various chaotic systems. Finally, the method is applied to some real world data to demonstrate the general relationship between the onset and progression of faults and changes in the largest Lyapunov exponent of a nonlinear system.Item Pathways, Networks and Therapy: A Boolean Approach to Systems Biology(2012-07-16) Layek, RitwikThe area of systems biology evolved in an attempt to introduce mathematical systems theory principles in biology. Although we believe that all biological processes are essentially chemical reactions, describing those using precise mathematical rules is not easy, primarily due to the complexity and enormity of biological systems. Here we introduce a formal approach for modeling biological dynamical relationships and diseases such as cancer. The immediate motivation behind this research is the urgency to find a practicable cure of cancer, the emperor of all maladies. Unlike other deadly endemic diseases such as plague, dengue and AIDS, cancer is characteristically heterogenic and hence requires a closer look into the genesis of the disease. The actual cause of cancer lies within our physiology. The process of cell division holds the clue to unravel the mysteries surrounding this disease. In normal scenario, all control mechanisms work in tandem and cell divides only when the division is required, for instance, to heal a wound platelet derived growth factor triggers cell division. The control mechanism is tightly regulated by several biochemical interactions commonly known as signal transduction pathways. However, from mathematical point of view, these pathways are marginal in nature and unable to cope with the multi-variability of a heterogenic disease like cancer. The present research is possibly one first attempt towards unraveling the mysteries surrounding the dynamics of a proliferating cell. A novel yet simple methodology is developed to bring all the marginal knowledge of the signaling pathways together to form the simplest mathematical abstract known as the Boolean Network. The malfunctioning in the cell by genetic mutations is formally modeled as stuck-at faults in the underlying Network. Finally a mathematical methodology is discovered to optimally find out the possible best combination drug therapy which can drive the cell from an undesirable condition of proliferation to a desirable condition of quiescence or apoptosis. Although, the complete biological validation was beyond the scope of the current research, the process of in-vitro validation has been already initiated by our collaborators. Once validated, this research will lead to a bright future in the field on personalized cancer therapy.Item Use of Autoassociative Neural Networks for Sensor Diagnostics(Texas A&M University, 2005-02-17) Najafi, MassiehThe 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.