Damage Identification in Wind Turbine Blades 2 nd Annual Blade Inspection, Damage and Repair Forum, 2014 Martin Dalgaard Ulriksen Research Assistant, Aalborg University, Denmark
Damage Identification in Wind Turbine Blades
2nd Annual Blade Inspection, Damage and Repair Forum, 2014
Martin Dalgaard Ulriksen Research Assistant, Aalborg University, Denmark
Presentation outline
• Research motivation
• Basic principles of damage identification
– Identification levels
– Physical quantities typically used
• Vibration-based damage identification
– Measurement of vibrations
– Applicable vibration quantities
• Case study
• Conclusions
2
Research motivation
Reliable damage identification enables, i.a., the turbine operators to:
• optimize maintenance
• shut down in case of an
emergency
3
Research motivation - continued
Cracks Edge damages Surface and
coating damages
4
Cracks and edge debondings are most critical damage types - require structural repairs.
Basic principles of damage identification
As defined by A. Rytter, damage identification covers 4 accumulative steps:
1. Damage detection
2. Damage localization
3. Damage assessment
4. Damage consequence
Example with damage length L:
5
Lvl. 2 Lvl. 3 Lvl. 2
Basic principles of damage identification – cont.
Quantities typically used for damage identification:
• Temperature
• Noise
• Vibration
6
Basic principles of damage identification – cont.
7
Temperature-based (thermography)
Basic idea: use infrared thermography to detect subsurface anomalies on the basis of temperature differences on the investigated surface.
• Advantages:
• Characterization of stress distributions and identification of stress concentration areas of
a surface • Area investigating technique
• Disadvantages: • Sensitivity towards spatial and temporal
temperature variations • Local measurements to assess damages
Basic principles of damage identification – cont.
Noise-based (acoustic emission)
Basic idea: monitor the acoustic emission generated by onset or growth of damage.
8
• Advantages: • Identifying damage areas plus hot spots and weak points
• Disadvantages: • Relatively high acoustic energy
attenuation (diversity of materials)
Basic principles of damage identification – cont.
Vibration-based
Basic idea: monitor the vibrations and examine signal anomalies.
9
• Advantages: • Independent of structural material
• Disadvantages: • Sensitivity difference in modal parameters
for different damage types
Basic principles of damage identification – cont.
Applicability of different methods for damage identification: Damage types: 1) Cracks, 2) Edge damages, 3) Surface and coating damages
10
Vibration-based damage identification
Vibrations can be measured as, e.g., displacements, velocities, and accelerations. Common for wind turbines is to mount wire-less accelerometers.
Based on time-dependent accelerations, the so-called modal parameters can be extracted through Operational Modal Analysis (OMA).
• Eigenfrequencies
• Mode shapes
• Damping ratios (not suitable for damage identification)
11
Vibration-based damage identification – cont.
• Eigenfrequencies (global parameter): – Natural frequencies of vibration for a system. Depends on the relation
between stiffness and mass of the system.
• Mode shapes (local parameter): – Relative motion between degrees of freedom when vibrating at
eigenfrequencies.
Beam system 1. mode 2. mode
12
Vibration-based damage identification – cont.
Numerous damage identification methods utilizing eigen-frequencies and/or mode shapes have been proposed.
First, we examine methods based on direct comparison between pre- and post-damage eigenfrequencies and mode shapes to see why these are inapplicable. Subsequently, we look at a more sophisticated mode shape-based method.
13
Case study
Damage identification in SSP 34 m wind turbine blade.
14
Case study – continued
Measurements during approximately seven minutes, corresponding to at least 500 oscillations at the lowest frequency of interest (≈ 1.3 Hz).
15
Only one cable for 1. Data 2. Synchronization 3. Power supply
Short accelerometer cable
Tri-axial accelero-meter mounted on swivel base
Case study – continued
Introduction of a 1.2 m trailing edge debonding (3.5 % of blade length) by use of hammer and chisel. The debonding was initiated 18.8 m from the blade root.
16
Case study – continued
Excited by hits with foam-wrapped wooden sticks at several locations along the blade (simulating ambient vibrations).
17
Case study – continued
18
OMA setup: • Unmeasured input: hits with foam-wrapped wooden sticks. • Measured output: accelerations in 20 points.
1.2 m debonding
Case study – continued
Eigenfrequency findings:
19
Natural frequencies, Hz
Diff.,% Undamaged Damaged
Mode Name Mean Confid.,% Mean Confid.,% 1 1st flap 1.36 0.79% 1.35 0.55% 0.48% 2 1st edge 1.86 0.47% 1.86 0.28% -0.10% 3 2nd flap 4.21 0.09% 4.21 0.16% 0.09% 4 2nd edge 7.12 0.04% 7.12 0.12% 0.11% 5 3rd flap 9.19 0.64% 9.17 0.13% 0.18% 6 1st torsion 12.40 0.18% 12.37 0.11% 0.24% 7 4th flap + 3rd edge 14.99 0.10% 14.98 0.09% 0.10%
The difference is much smaller than
the confidence!
Case study – continued
Mode shape findings:
• No traces of the damage at the lowest modes
20
1st flapwise mode 1st edgewise mode
Case study – continued
Mode shape findings:
• No traces of the damage at the lowest modes
• Some differences occur in the higher modes
21
8th mode (combination of flap and edge)
Case study – continued
Direct comparisons of pre- and post-damage modal parameters do not facilitate valid damage identification. Therefore, continuous wavelet transformation (CWT) is employed.
CWT: Calculates similarity between a signal and a so-called wavelet function. Works as a discontinuity/irregularity scanner.
22
Case study – continued
CWT results by use of 8th mode (combination of 3rd edgewise and 4th flapwise bending modes) and a 4th order Gaussian wavelet:
(a) CWT of post-damage signal-processed 8th mode shape. (b) CWT of pre-damage signal-processed 8th mode shape. (c) Difference between (a) and (b).
23
Case study – continued
The CWT plotted in Fig. c in the previous slide is converted to a simple statistical damage indicator. States 1-4 are damaged, while states 5-8 are undamaged.
24
Statistical threshold: above = no damage below = damage
Conclusions
• Modal parameters of the lower modes are not the best indicators of a damage.
• For damage localization and especially assessment, known methods are highly dependent on the number of measurement points (e.g. number of accelerometers).
• Wavelet transformation shows potential for damage identification in wind turbine blades.
• A study on the general applicability of the method is necessary. The study includes, i.a.: – Tests with rotating blade (full operational condition).
– Measurement point density.
25
Thank you for your attention.
26