Detection of Damage in OperatingWind Turbines by Signature Distances

Abstract

Wind turbines operate in the atmospheric boundary layer and are subject to complex random loading. This precludes using a deterministic response of healthy turbines as the baseline for identifying the effect of damage on the measured response of operating turbines. In the absence of such a deterministic response, the stochastic dynamic response of the tower to a shutdown maneuver is found to be affected distinctively by damage in contrast to wind. Such a dynamic response, however, cannot be established for the blades. As an alternative, the estimate of blade damage is sought through its effect on the third or fourth modal frequency, each found to be mostly unaffected by wind. To discern the effect of damage from the wind effect on these responses, a unied method of damage detection is introduced that accommodates different responses. In this method, the dynamic responses are transformed to surfaces via continuous wavelet transforms to accentuate the effect of wind or damage on the dynamic response. Regions of signicant deviations between these surfaces are then isolated in their corresponding planes to capture the change signatures. The image distances between these change signatures are shown to produce consistent estimates of damage for both the tower and the blades in presence of varying wind eld proles.