Blade faults diagnosis in multi stage rotor system by means of wavelet analysis

Blade fault is one of the most causes of gas turbine failures. Vibration spectral analysis and blade pass frequency (BPF) monitoring are the most widely used methods for blade fault diagnosis. These methods however have limitations in the detection of incipient faults due to weak and/or transient si...

Full description

Saved in:
Bibliographic Details
Main Author: Ahmed, Ahmed Mohammed Abdelrhman
Format: Thesis
Language:English
Published: 2014
Subjects:
Online Access:http://eprints.utm.my/id/eprint/77817/1/AhmedMohammedAbdelrhmanPFKM2014.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Blade fault is one of the most causes of gas turbine failures. Vibration spectral analysis and blade pass frequency (BPF) monitoring are the most widely used methods for blade fault diagnosis. These methods however have limitations in the detection of incipient faults due to weak and/or transient signals, as well as inability to diagnose the blade faults types. This study investigates the applications of wavelet analysis in blade fault diagnosis of a multi stage rotor system, as an extension of previous works which involved a single stage only. Results showed that conventional wavelet analysis has limitations in segregating the BPFs and locating the faults. An improvement in Morlet wavelet was made to achieve high resolution in both time and frequency domains. Two new wavelets for high time-frequency resolutions were formulated and added to the standard MATLAB Wavelet Toolbox. The optimal parameters for the high frequency resolution wavelet were found at the centre of frequency, ????=4 and bandwidth, ??=0.5. For high time resolution wavelet, the optimal parameters were ????=4 and ??=10. A novel algorithm was formulated by combining the two newly developed wavelets. A variety of blade faults including blade creep rubbing, blade tip rubbing, stage rubbing, blade loss of part and blade twisting were tested and their vibration responses measured in a laboratory test facility. The proposed method showed potential in segregating closely spaced BPFs components and identifying the faulty stage and fault location. The method demonstrated the ability in differentiating various blade faults based on a unique pattern (“fingerprint”) of each fault produced by the newly added wavelet. The formulated algorithm was demonstrated to be suitable in monitoring rotor systems with multiple blade stages.