Research On Fault Diagnosis Of Wind Turbine Gearbox

At present, the energy problem and environmental pollution has become an increasingly important issue, and the energy crisis gets worse, therefore, the concern of new energy gradually rises. Wind energy is inexhaustible and renewable green energy with the advantages of no pollution, no noise, no waste. And the wind power has the features of the short investment cycle and less occupation of the land, development of new energy has gradually attracts more focus and attention to the world. With the rapid increasement in installed capacity of large-scale wind turbine in recent years, and the system structure is more complexed, so the reliability and stability of operation is highly required. When the fault o f wind turbine occurs, it will not only cause power outage, but also result in serious accidents and the huge economic losses.Since the gearbox failure frequently occurs in all components of wind turbine, the downtime of the gearbox failure is longest. The gearbox fault will cause the greatest loss of generating capacity, so the gearbox is taken as the research object. The following is the main work and innovative research of this paper:Firstly, HHT method is used in fault feature extraction of rolling bearing in this paper, HHT is an adaptive signal process method to cope with non-stationary signal, it will decompose a arbitrarily signal into a finite number of intrinsic mode functions. And we can use Hilbert transform to get the instantaneous amplitude and frequency of the IMFs. The Hilbert spectrum represents the relationship between frequency and time of the signal. In the decomposition process of EMD, since the endpoints of the signal may not be the extreme points, these will cause the end effect and mode mixing problem. EEMD is used to solve the mode mixing problem due to the addition of white gaussian noise, then mutual information method is adopted to eliminate the false components produced by the decomposition.Given the special nature of wind turbine vibration signals, especially the severe work environment and the noise interference, which seriously effects the extraction and analysis of wind turbine vibration signals. When the fault is in the early stages, strong background noise interferes with the effective recognition of the wind turbine condition. Aiming at the problems of wind turbine harsh work environment and the severe noise pollution, a de-noise method based on mathematical morphology filter is applied in the pre-treatment of wind turbine vibration. Compared with traditional vibration signal processing method, this method in dealing with signal is entirely in the time domain, and the calculation is easy and fast.Therefore, an extension method based on the grey prediction theory is used to suppress the end effect problem. The improved method can be effectively applied to fault diagnosis of rolling bearings according to simulation results.