Turbine blade crack detection using vibration testing methods

摘要

Turbine blades are the most common cause of failures in gas turbines. Failure modes are typically cracking from foreign object damage (FOD), high cycle stress (HCF), blade rubbing, degradation from erosion and corrosion. Fault detection of blades is important function in reducing blade related failures. This study involved the use of vibration analysis and dynamic testing of blades for failure detection. Current field inspections of blades are based on visual inspections only, and the intent here was to use impact testing of the blades during these inspections to determine cracked blade (from the vibration response). Vibration impact tests were undertaken using decommissioned turbine blades with and without cracks in the laboratory. Four common crack patterns were deliberately induced to the turbine blades to investigate changes in the blades normal mode response. Finite element analysis (FEA) of the blades was also undertaken. FEA results were correlated to experimental results and these results showed that each crack pattern was unique and significant changes were found in higher modes. Dynamic vibration analysis was also undertaken in a laboratory testrig fitted with rotating model straight blades. Vibration measurements were undertaken on a baseline test case and other test conditions where the blades had induced cracks. Steady state and transient vibration responses were obtained from the controlled tests. Vibration measurements were on the machine casing and on a blade (with an accelerometer surfaced mounted onto a rotating blade). Order tracking analysis and continuous wavelet analysis were performed to the measured data. Both techniques showed ability in detecting changes in vibration response on the blades and casing for blades with cracks. Results showed that significant changes were detected during transient analysis as compared to the steady state. Changes were noted in the vibration response corresponding to the machine running speed and blade passing frequencies. Changes in wavelet maps were however qualitative in nature, which suggested that more work need to be undertaken for cracks severity assessment using wavelet maps.