Kalman filter-based artificial neural networks for wind turbines remaining useful life prediction

Abstract

High-speed shaft bearing (HSSB) failure in wind turbines (WTs) is a major problem that can cause unexpected blackouts in the electricity generation. Research on naturally occurring, advanced faults in HSSBs is relatively scarce, and online damage severity assessments are rarely reported in published works. In this paper, two approaches are outlined to predict degradation in HSSBs (Kalman filter (KF) and artificial neural networks (ANNs)), specifically inner race failure, and to estimate the remaining useful life (RUL) of HSSB. The health indicators (HIs) are used to evaluate the degradation of HSSB, and the minimum redundancy maximum relevance (mRMR) method is used to extract the best suitable HI for HSSB prognosis purpose. Therefore, a single prognostic metric is employed to evaluate the RUL prediction outcomes. To validate the accuracy of the methods, experiments were conducted on an actual HSSB vibrations dataset. Comparing with current state-of-the-art methods on the same dataset demonstrates the performance of the proposed methods.