National Technical University of Athens (NTUA)
Εφαρμογή νοημόνων μεθόδων επεξεργασίας δυναμικών σημάτων στη διάγνωση βλαβών μηχανολογικού εξοπλισμού
Abstract
dc:descriptionThe demodulation of dynamic signals is studied using two approaches towards the aim of the automated export of the diagnostic information. The first approach is based on the Teager-Kaiser non-linear Differential Energy Operator and the Energy Operator Separation Algorithm – EOSA is used. The second approach is based on the Complex Shifted Morlet Wavelets. A number of methodologies and criteria have been proposed for the automatic selection of the two optimal wavelet parameters. These criteria are systematically analyzed and compared. A new criterion for the simultaneous optimal selection of the critical parameters is proposed, offering opportunities in order to automate the extraction of the envelope of the signal and detect most reliably and safest the demodulation band. Moreover, the detection of cracks in beam structures is attempted using a combination of the Principal Orthogonal Decomposition and the Morphological Processing. The four basic morphological operators are compared, using different types of structuring elements. The operator of erosion (erosion) with a curved (spline) structuring element presents the better results. The performance of the method is assessed using different beam structures that present a "breathing" crack and several factors, such as the boundary conditions, the location of the crack, the depth of the crack, the existence of multiple cracks, the distance between the multiple cracks, the level of noise in the measurements and the spatial resolution of the measured points. Furthermore, an automatic system for the fault diagnosis of ball element bearings is proposed, using the method of Support Vector Machines. The automatic fault diagnosis can be divided into two main parts, the input – preprocessing of the signals and the classification procedure. A hybrid algorithm which is based on a stochastic simulation model of the dynamic response of ball bearings, on the strategy of One-Against-All and on the method of Support Vector Machines is proposed. The simulated signals are generated by a well defined model that describes the dynamic response of damaged bearings. The features, which can be extracted from the spectra of both the raw signal and the envelope of the signal and characterize the changing condition of the engine, are considered. The key aspect is the extraction of the energy which is developed and spread at carefully, correctly and precisely defined frequency bands, that can be determined based on the good understanding of the nature of the phenomenon of the occurrence of the damage from an engineering standpoint. The successful definition of the frequency bands can lead to a more accurate, reliable and automated monitoring of rotating machinery and to accurate and on time fault diagnosis. Likewise in order to overcome the deterioration of the spectrum due to the emergence of transient phenomena, the use of Order Analysis is proposed. The proposed methods are evaluated using simulated and real signals. The real signals are derived from measurements performed at industrial plants and from international databases containing experimental measurements.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2009
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Gryllias, Konstantinos
- Γρύλλιας, Κωνσταντίνος
Subjects
dc:subject × 20- Εντοπισμός & διάγνωση βλαβών
- Ανάλυση κραδασμών
- Ένσφαιροι τριβείς
- Μαθηματική μορφολογία
- Εντοπισμός ρωγμών
- Μηχανές διανυσματικής υποστήριξης
- Κυματίδια
- Μέθοδοι αποδιαμόρφωσης
- Fault detection & diagnosis
- Vibration analysis
- Rolling element bearings
- Mathematical morphology
- Crack detection
- Support vector machines
- Wavelets
- Demodulation methods
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Μηχανολόγου Μηχανικού
- Engineering and Technology
- Mechanical Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/22344