Statistical distribution measures based on amplitude normalization for wind turbine generator bearing condition monitoring under variable speed conditions

Abstract

Wind turbines (WTs), with the capacity of renewable energy production, have been massively equipped in recent years. To improve the reliability of the WTs and also reduce the operation and maintenance (O&M) costs, condition monitoring based preventative maintenance is of urgent need. For this industrial application demand, health indicator (HI) construction is a promising solution. However, it should be noted that most of the currently available HIs are developed based on the assumption of stationary or quasi-stationary operating conditions, the performances of which in time-varying speed cases, nevertheless, are significantly influenced due to the dynamic interactions. Aiming at this issue, a statistically interpretable HI based on the amplitude normalization is proposed in this paper. In this method, an amplitude normalization strategy is firstly designed to suppress the variable speed induced interferences. Afterwards, a characteristic model is established for the integrated statistical representation of the signal from the distribution perspective. Multiple parameters in this model are estimated by the maximum log-likelihood method. Then the evolution of the established probability distribution during the degradation process is analyzed, the statistic deviation is accordingly estimated and taken as a novel HI to characterize the degradation process of the WT generator bearing. Finally, with the simulated bearing degradation data and the industrial field datasets collected from different WT generator bearings, experimental tests are conducted and indicate that the proposed method is preferable in bearing degradation process characterization under variable speed conditions.