EXPERIENCE WITH BICOHERENCE OF ELECTRICAL-POWER FOR CONDITION MONITORING OF WIND TURBINE-BLADES

The authors explore the application of the normalised bispectrum or bi coherence to the problem of condition monitoring of wind turbine blade s. Background information is provided on this type of condition monito ring, how it differs from more conventional condition monitoring of tu rbo machinery, and the motivation for selecting bicoherence. Bicoheren ce is defined and compared with the power spectral density. Complicati ons in collecting suitable data, and estimating the bicoherence from t hat data are investigated; including the requirements of very long sta tionary data sets for consistent estimates, and computational difficul ties in handling such large data sets. Bicoherence is then applied to electrical power output data obtained from a 45kW wind turbine. The tu rbine is operated in three configurations to represent normal and faul t conditions. A blade with less flapwise stiffness but identical outer dimensions to the matched set of blades was fitted to simulate a dama ged blade. Comparison of the results from the power spectral density a nd bicoherence indicates how the bicoherence might be employed for con dition monitoring purposes. Slices of the bicoherence with one frequen cy fixed at the rate of rotation show clear differences between the co nfigurations and substantially reduce the computational effort require d to calculate the estimate.