Research On Fault Trend Prediction Method Of Wind Turbine Gearbox

Gearbox is the key equipment during the operation and maintenance of doubly-fed wind turbines, and not only has a high failure rate, but needs relatively longer repair time once a fault occurs, which can lead to huge economic losses.This paper conducts research on fault trend prediction method of wind turbine gearbox, aiming to judge the future running state of gearbox, predict its development trend and estimate the remaining life, which will provide important reference and basis for the operation and maintenance of wind turbines.Firstly, this paper gives a brief overview about the common structures of the wind turbine gearbox, and then the typical fault types of gearbox are described based on four kinds of the most common failure modes (fatigue mode, wear mode, overload mode and corrosion mode). Meanwhile, signal of typical gear fault is simulated according to the gear vibration mechanism.Secondly, based on prediction theory and hypothesis of fault trend theory, this paper proposes the fault trend prediction process of wind turbine gearbox, which is classified into two prediction ways:based on the data and based on the fault model, and then the process steps are introduced briefly.Thirdly, this paper takes a group of fault characteristic values as the forecasting object, and achieves long-term trend prediction with the application of time series analysis method, BP neural network prediction model and grey prediction model. Then, the remaining life is estimated by setting the fault threshold of the fault characteristic value. At the same time, taking the real fault signal as an example, a comparison of different prediction methods is made with the analysis and improvement of different prediction models.Finally, taking the fatigue mode as an example, this paper conducts some research on the fault trend prediction method based on the fault model. According to the Miner linear cumulative damage theory, the gear damage is calculated by application of load spectrum analysis method. Then, taking the gear daily cumulative damage level over a period of time as basis data, the daily cumulative damage for a period of time in the future is predicted, and the residual life is also estimated, which validates the feasibility and validity of the combination of data prediction with the fault model.