Improved Blade Fault Diagnosis Using Discrete Blade Passing Energy Packet and Rotor Dynamics Wavelet Analysis

Abstract

Blade fault represents one of the most frequent causes of gas turbine failures. Although various measurement methods (i.e. pressure, strain gauges, and blade tip measurements) have been found to be effective in diagnosing blade faults, it is often difficult to deploy these methods under field conditions due to the requirement of mounting sensors in the interior of a running gas turbine. Vibration spectra analysis is inevitably still represents the most widely used method for blade fault diagnosis under field conditions. However, this method is known to be only effective in detecting severe blade fault conditions (i.e. terminal rubbing); whilst, minor and transient blade faults (i.e. geometry alterations, reduction in blade tip clearance, and Foreign Object Damage (FOD) event) are often left undetected. This makes vibration spectra analysis an unreliable tool for total blade fault diagnosis in the field. This study was thus conducted to investigate methods that can improve the sensitivity and reliability of vibration analysis for blade faults diagnosis. Two novel vibration analysis methods were formulated, namely the Rotor Dynamic Wavelet Map (RDWM) and Blade Passing Energy Packet (BPEP). Experimental results showed that the time-frequency display of RDWM could provide a clearer picture of the rotor dynamic characteristics of a rotor system compared to vibration spectra. RDWM also provides a better visualization of the blade condition in the rotor and enables discrimination of various blade fault conditions (i.e. creep rub and eccentricity rub). Meanwhile, the BPEP method which breaks the overall Blade Passing Frequency (BPF) component into instantaneous and discrete energy packets of running blades in the rotor system, enables a more sensitive detection of rotor eccentricity conditions and provides early warning for impending blade rubbing which is often undetectable in the vibration spectra.