In the present report, modal and wavelet analysis-based damage identification for wind turbine
blades is addressed. Damage identification is typically classified by use of four levels, namely
1) detection, 2) localization, 3) assessment and 4) consequence. Since this report is the first of
its kind at Aalborg University Esbjerg, it is chosen only to treat level 1, 2 and 3 identification,
because these levels compose an appropriate basis for conducting fundamental research within
the area of damage identification and subsequent expansion to more specific and complex issues.
A literature review of some of the general methods available for damage identification is
presented along with a description of typical structural damages in wind turbine blades. On
the basis of the review and damage descriptions, a modal analysis-based approach is preferred
since this method facilitates up to level 3 identification and shows promise when it comes to
in-service employment.
With modal analysis as the general approach, the damage identification must be conducted
by use of a specific method which utilizes modal parameters, i.e. eigenfrequencies, mode shapes
and modal damping ratios, as damage indicators. Both the eigenfrequencies and the damping
ratios have proven to be inconvenient, because they, respectively, are too insensitive towards
structural damages and show ambiguous changes when damage is introduced. Consequently,
three methods utilizing mode shapes or their derivatives have been tested on the basis of
numerically derived pre- and post-damage mode shapes of a cantilevered rectangular shell
tube. Hereby, it is found that a wavelet analysis-approach yields the best results for damage
detection, localization and assessment, and therefore this method is chosen.
To examine the general applicability of the modal and wavelet analysis-based damage iden-
tification method, blades from a residential-sized wind turbine, namely the Whisper 500, and
a full-scale wind turbine, namely the GE 1.5 XLE, are treated. first, level 1, 2 and 3 dam-
age identification is conducted for the Whisper 500 blade introduced to longitudinal cracks of
realistic sizes. The mode shapes are derived both experimentally and numerically, and since
the different results show strong correlation, the experimental and numerical findings show
the same tendencies; introduction of longitudinal cracks mostly influences the mode shapes of
the higher modes, especially the twisting modes, but through wavelet analysis even the first
bending mode shape yields valid and useful damage detection, localization and assessment.
Damage identification analyses of the GE 1.5 XLE wind turbine blade are conducted for
derived mode shapes of a numerical blade model, to which different damage types, in the form
of longitudinal and transverse cracks plus edge debondings, are introduced both alternately and
simultaneously. To further challenge the robustness and general applicability of the identifica-
tion method, Gaussian noise is added in order to simulate typical experimental findings. It is
found that valid damage identification is obtained for both single- and multi-damage scenarios
with noise-contamination corresponding to a signal-to-noise ratio (SNR) of 40 dB when the
second flapwise bending mode and the first twisting mode are employed along with the first
flapwise bending mode.