| Summary: | Reliable and efficient condition monitoring (CM) techniques play a crucial role in minimising wind turbine (WT) operations and maintenance (O&M) costs for a competitive development of wind energy, especially offshore. Although all new turbines are now fitted with some form of condition monitoring system (CMS), very few operators make use of the available monitoring information for maintenance purposes because of the volume and the complexity of the data. This Thesis is concerned with the development of advanced automatic fault detection techniques so that high on-line diagnostic accuracy for important WT drive train mechanical and electrical CM signals is achieved. Experimental work on small scale WT test rigs is described. Seeded fault tests were performed to investigate gear tooth damage, rotor electrical asymmetry and generator bearing failures. Test rig data were processed by using commercial WT CMSs. Based on the experimental evidence, three algorithms were proposed to aid in the automatic damage detection and diagnosis during WT non-stationary load and speed operating conditions. Uncertainty involved in analysing CM signals with field fitted equipment was reduced, and enhanced detection sensitivity was achieved, by identifying and collating characteristic fault frequencies in CM signals which could be tracked as the WT speed varies. The performance of the gearbox algorithm was validated against datasets of a full-size WT gearbox, that had sustained gear damage, from the National Renewable Energy Laboratory (NREL) WT Gearbox Condition Monitoring Round Robin project. The fault detection sensitivity of the proposed algorithms was assessed and quantified leading to conclusions about their applicability to operating WTs.
|
|---|
