Fault diagnosis of wind turbine gearbox based on Fisher criterion

The fault and normal gearboxes are classified and identified by principal components analysis (PCA) and Fisher criterion based on the vibration test and time-domain analysis for several wind turbine gearboxes. First the multi-dimension time-domain feature values are extracted from the gearbox vibration signals and PCA is carried out for dimension compression. Then the feature data which the dimension is reduced are classified and identified by Fisher criterion and the classification threshold is given. The result shows that the sensitivities of different time-domain feature values for the gearbox fault are different. But the differences information of the original feature space is preserved while the dimension is reduced by PCA. The classification for the reduced-dimension feature space is succeeded in the Fisher criterion.