Bladed Disk Crack Detection through Advanced Analysis of Blade Passage Signals

By: Elhamosadat Alavi Foumani

Ottawa- Carleton Institute for Mechanical Engineering Ottawa, Canada Fall 2012

Outline
Introduction Measurements and Experimental set up Tip Clearance and Time of Arrival Preprocessing Analysis Statistical Wavelet Analysis Conclusions Future works Acknowledgement

http://usatoday30.usatoday.com

Click View then Header and Footer to change this footer

Source: http://www.bbc.co.uk/news/world-africa-18831686

Introduction
Blade position measurements Spin pit test facility Apparatus for predicting failures of spinning disks Input Data: Tip clearance, and Time of Arrival

Measurements Scheme
Lengthened blades Crack in hub

Blade length on healthy hub (including temperature and RPM variation)

Asymmetric deformation of a turbo machinery disk due to crack propagation at the base of a set of blades

* A. Von Flotow, "Apparatus and method for predicting failures of spinning disks in turbo-machinery." U.S. Patent No. 6,785,635. 31 Aug. 2004.

Measurements
Blade length near hub crack
Elastic deformation increases with decreasing hub stiffness

Blade length on healthy hub

Plastic deformation: Blade does not retract during spin-down cycle

A general illustration ofin bladeresponse and of identifycycled.deformation. Changes sensor length as a turbo disk is blade

* A. Von Flotow, "Apparatus and method for predicting failures of spinning disks in turbo-machinery." U.S. Patent No. 6,785,635. 31 Aug. 2004.

Tip Clearance Signal Noise Removal

Temperature effect Elastic stretch due to centrifugal force
Demean TC signal based on cycles

Disk Imbalance effect

Fourier Series

1 = 0 + 2
=1

cos + sin
2

1 =

cos
=1

2 1 =

=1

2 sin

Disk Imbalance Sinusoidal effect

Magnitude 1 = 1 2 + 1 2

Phase 1 = tan1 ( 1 )
1

The sinusoidal effects carried by the TC signal and caused by the rotor imbalance: (a) the sinusoidal effects over 1408 cycles; (b) the magnitude; (c) the phase.

Tip Clearance Preprocessing Analysis

(a) The localized standard deviation trend of the magnitudes of the sinusoidal effects carried by the TC signal; and (b) the localized standard deviation trend of the phases.

Time of Arrival Preprocessing Analysis

dt in Time ) = dx / is an interpretation Deviation= dy / (R of Arrival(R tan ) of axial displacement of a blade

10

Time of Arrival Preprocessing Analysis

dR

Angular deformation of rotor shaft causes a harmonic term in Time of Arrival The shaft bending due to eccentricity which causes an Angular Deflection and a Radial Displacement dR

11

Time of Arrival Preprocessing Analysis

The sinusoidal effects carried by the ToA signal and caused by the shaft bending: (a) the sinusoidal effects over 1408 cycles; (b) the magnitude; and (c) the phase;

12

Time of Arrival Preprocessing Analysis

The sinusoidal effects carried by the ToA signal and caused by the shaft bending: (a) the localized standard deviation trend of the magnitudes; and (b) the localized standard deviation trend of the phases.

13

Digital Signal processing Techniques

Time Domain analysis Frequency domain Analysis Time-Frequency Domain analysis Wavelet Transform Analysis Daubechies 4
+

, =

, , 0

14

Statistical Wavelet Analysis

STD of WT Coef

0 60 40 20 10 Scale # 0 0 Blade # 20 40 30

The standard deviation of the wavelet coefficients over 48 scales for 40 blades based on the Time-of-arrival signal.

15

Statistical Wavelet Analysis

The standard deviation of the wavelet coefficients over 48 scales for 40 blades based on the Tip Clearance signal.

16

Statistical Wavelet Analysis

The STD of the wavelet coefficients over 48 scales for 1408 cycles based on the Tip Clearance signal.

17

Conclusions
Having proper denoising prior to analyses, application of the wavelet method for fault detection in bladed discs found to be promising. Results show that the statistical wavelet analysis technique is capable of providing useful indicative information for locating the incipient crack that occurs in the bladed disk

The crack initiation time can be determined by using the suggested localized standard deviation of the sinusoidal effects caused by the rotor imbalance

It was found that the physical based modeling, can effectively characterized the noises induced by various known factors other than defects such as thermal and centrifugal expansions, rotor imbalance, and shaft bending effects. These noises

subsequently removed.

19

Future works
The present work is done based on the Tip clearance and Time of arrival data provided by LPT Inc. The available data may not allow more comprehensive analysis. If

additional data are available in future, further studies in the following directions could
be worth exploring: Identifying the various stages of damage evolution during its life cycle by extracting other effective parameters except standard deviation and kurtosis from

wavelet map.
Different wavelets be examined to develop efficient algorithms to identify the stress levels, crack initiation, crack growth to its final rupture. The characteristics of the modulation change during the life cycle of the disk spin rig test should be examined.

20

Acknowledgement

I would like to acknowledge people who made this work possible: Prof. Ming Liang, Dept. of Mechanical Engineering, University of Ottawa Prof. Jie Liu, Dept. of Mechanical and Aerospace Engineering,

Carleton University
Life Prediction Technologies Inc. (LPTi)

21

Thank you for your attention

?
22