Unsupervised Pattern Recognition of Acoustic Emission from Full Scale Testing of a Wind Turbine Blade


J. of Acoustic Emission, Vol. 18, 2000, pp 217-223

(Initially presented and published at the proceedings of EWGAE 2000 - 24th European Conference on AE Testing, CETIM - France, 24-26 May, 2000, pp. 291-297)

D. Kouroussis, A. Anastasopoulos, P. Vionis, V. Kolovos


Acoustic Emission (AE) monitoring during full scale testing of FRP Wind Turbine (W/T) blades is a, relatively, new application. The difficulty in such tests arises from the potentiality of different AE sources expected, due to the nature of FRP materials, as well as the complex design of W/T blades. AE data obtained during a static proof test of a 12m FRP Wind Turbine blade was analyzed, in order to assess the criticality of specific AE sources. Unsupervised pattern recognition (UPR) was used to segregate the AE signals into various classes, based on the similarity of AE features. The yielded classes were, then, correlated with the applied load, and the AE characteristics of each class were compared. Particular classes were observed to appear at early load stages, but ceased at higher loads, while some classes were considerably active during high loads and load-holds. The application of UPR on such AE data was proved to be a powerful tool towards the characterization of damage evolution with increasing load. Combined use of UPR and location of particular AE sources, located with traditional location algorithms was investigated, revealing that a localized source can yield more than one classes as the damage mechanism associated with it might change in time and with respect to the applied load.