Condition monitoring algorithm suitable for wind turbine use

Wind turbine condition monitoring is gaining importance as turbines increase in size and move further offshore. Monitoring signals from turbine drive trains are highly non stationary as turbine load and speed vary with time. Manual analysis of signals from multiple turbines is challenging so an automated algorithm is required. Monitoring signals contain fault frequencies of interest that are speed dependent and so vary with time. The frequency tracking methodology allows these components to be tracked and their amplitude plotted over time without the need for manual analysis of multiple wide-band spectra. Since wind turbines are inherently sinusoidal, a frequency tracking algorithm based on Fourier principles is proposed and discussed. The algorithm is successfully applied to detect changes in fault level in electrical and mechanical monitoring signals on a variable speed laboratory test rig.