Prognosticating fault development rate in wind turbine generator bearings using local trend models

Generator bearing defects, e.g. ball, inner and outer race defects, are ranked among the most frequent mechanical failures encountered in wind turbines. Diagnosis and prognosis of bearing faults can be successfully implemented using vibration based condition monitoring systems, where tracking and trending of specific condition indicators can be used to evaluate the former, current and potentially future condition of these components. The latter, i.e. evaluation of the fault progression rate and remaining useful lifetime (RUL), is of essential importance to owners and operators in regards to maintenance planning and component replacement. The above approach offers numerous benefits from financial and
operational perspective, such as increased availability, uptower repairs and minimization of secondary and catastrophic damages. In this work, a non-speed related condition indicator, measuring the signal energy between 10Hz to 1000Hz is utilized as feature to characterize the severity of developing bearing faults. Furthermore, local trend models are employed to predict the progression of bearing defects from a vibration standpoint in accordance with the limits suggested in ISO 10816. Predictions of vibration trends from multi-megawatt wind turbine generators are presented, showing the effectiveness of the suggested approach on the calculation of the RUL and fault progression rate.