Fault diagnosis of wind turbine blades using histogram features through nested dichotomy classifiers

Abstract

This study makes an attempt of classifying different fault conditions which occurs on wind turbine blade due to environmental stress and high wind speed. “Here three bladed horizontal axis variable wind turbine was used for experimental study and different faults like blade crack, hub-blade loose connection, erosion, pitch angle twist and blade bend was considered. This study had been carried out in three phases namely feature extraction, feature selection and feature classification. Initially vibration signals are noted for different blade conditions and required features are obtained using histogram features. Secondly, from the extracted feature, most dominating feature need to be chosen using J48 decision tree classifier. Later, the selected feature is fed into the classifiers like Nested Dichotomy (ND), Class-Balanced Nested Dichotomies (CBND) and Data near Balanced Nested Dichotomy (DNBND) for classification of the faults. These classifiers are compared with respect to their accuracy to suggest a better model for fault diagnosis on blade. The suggested model can be incorporated in real-time system to monitor the condition of wind turbine blade.”.