Output-Only Techniques For Fault Detection.
dc.contributor.author | Brzezinski, Adam John | en_US |
dc.date.accessioned | 2011-09-15T17:15:00Z | |
dc.date.available | NO_RESTRICTION | en_US |
dc.date.available | 2011-09-15T17:15:00Z | |
dc.date.issued | 2011 | en_US |
dc.date.submitted | en_US | |
dc.identifier.uri | https://hdl.handle.net/2027.42/86429 | |
dc.description.abstract | Fault detection is relevant to many applications, including structural health monitoring and machine health monitoring. Furthermore, output measurement data may be the only information known about a system. Hence we develop and demonstrate techniques for output-only fault detection. We also investigate implementation issues, including computational complexity and output noise. First, we consider real-time detection of an abrupt change in a noisy signal. Existing techniques exhibit sensitivity to gradual (incipient) changes in the data, as well as detection delays, false alarms, and missed detections. Hence, we propose an adjacent moving window peak detection (AMWPD) approach that uses an approximate low-pass filter and statistical process control techniques to determine whether an abrupt change has occurred. We compare the AMWPD approach with existing techniques for change detection and show that the AMWPD approach exhibits comparable detection speed and number of missed detections while providing fewer false alarms. Second, we consider feature extraction and clustering for classification. For industrial applications, existing methods provide insufficient classification accuracy and require significant training time. Hence, we propose new features that improve classification accuracy and apply a modified tabu search and probabilistic neural network (mTS + PNN) approach to select and cluster the features and thereby classify the data. We compare the mTS + PNN approach with an existing feature selection and clustering technique that employs principal component analysis and a multi-layer perceptron neural network. Using an application example, we demonstrate that the mTS + PNN approach provides higher classification accuracy while requiring less training and classification time. Finally, we consider identification of output-to-output relationships in linear system dynamics. Existing approaches, including operational modal analysis, assume that the excitation signal is a realization of a white random process, which may not be true. Hence, we define and characterize pseudo transfer functions (PTFs), which relate output measurements, and we use changes in the identified PTF to detect faults. We demonstrate the effects of nonzero initial conditions, non-white excitation, unknown model order, and output noise on the accuracy of the identification and fault detection results. | en_US |
dc.language.iso | en_US | en_US |
dc.subject | Fault Detection | en_US |
dc.subject | Sensor-only | en_US |
dc.subject | Pseudo Transfer Function | en_US |
dc.subject | Health Monitoring | en_US |
dc.subject | Least Squares | en_US |
dc.subject | Feature Selection | en_US |
dc.title | Output-Only Techniques For Fault Detection. | en_US |
dc.type | Thesis | en_US |
dc.description.thesisdegreename | PhD | en_US |
dc.description.thesisdegreediscipline | Aerospace Engineering | en_US |
dc.description.thesisdegreegrantor | University of Michigan, Horace H. Rackham School of Graduate Studies | en_US |
dc.contributor.committeemember | Bernstein, Dennis S. | en_US |
dc.contributor.committeemember | Ni, Jun | en_US |
dc.contributor.committeemember | Atkins, Ella Marie | en_US |
dc.contributor.committeemember | Kolmanovsky, Ilya Vladimir | en_US |
dc.contributor.committeemember | Kukreja, Sunil L. | en_US |
dc.contributor.committeemember | Stein, Jeffrey L. | en_US |
dc.subject.hlbsecondlevel | Aerospace Engineering | en_US |
dc.subject.hlbsecondlevel | Civil and Environmental Engineering | en_US |
dc.subject.hlbsecondlevel | Engineering (General) | en_US |
dc.subject.hlbsecondlevel | Mechanical Engineering | en_US |
dc.subject.hlbsecondlevel | Science (General) | en_US |
dc.subject.hlbsecondlevel | Statistics and Numeric Data | en_US |
dc.subject.hlbtoplevel | Engineering | en_US |
dc.subject.hlbtoplevel | Science | en_US |
dc.description.bitstreamurl | http://deepblue.lib.umich.edu/bitstream/2027.42/86429/1/ajbrzezi_1.pdf | |
dc.owningcollname | Dissertations and Theses (Ph.D. and Master's) |
Files in this item
Remediation of Harmful Language
The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.
Accessibility
If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.