Feature fusion model based health indicator construction and self-constraint state-space estimator for remaining useful life prediction of bearings in wind turbines

Accurate remaining useful life prognosis of bearings in wind turbines is beneficial for operation and maintenance schedule at wind farms. The major technical challenges include how to construct the strictly monotonic and gently growing health indicators of bearings, and further improve the prediction performance under harsh operational environments. To deal with these challenges, this paper proposes a degraded feature fusion model aiming at the construction of monotonic health indicators. In the model, the features with higher monotonicity are selected as sensitive features, and used to calculate the degradation ratios for health indicators. A general-izable failure threshold determination method is presented to find the common characteristics of failure patterns of a set of indicators. Besides, a novel state-space estimator with self-constraint property is proposed, which can update the state-space in the future time for more robust remaining useful life prediction. Several sets of the bearings from PRONOSTIA platform and on-site wind turbine high-speed shafts are utilized to validate the proposed approach, the former results demonstrate that the prediction accuracy of the proposed approach is better than some existing algorithms, and the latter shows that the potential of the proposed approach to provide advices for maintenance schedule.