Fault Diagnosis of Wind Turbine Based on Empirical Mode Decomposition

The structure of the gearbox of wind turbine is complex. The working environment is bad and the load is heavy which results in frequent failure of the gearbox parts and huge economic losses. In this paper, the signal processing method combined with empirical mode decomposition and improved bispectrum is used to diagnose the gearbox of wind turbine and the condition of wind turbine is monitored by a portable inspection. The feasibility and applicability of this method in the fault diagnosis of wind turbine gearbox are checked by an example.


Introduction
Wind power has become a key research area in the world. Once the wind turbine fails, it will cause serious economic losses. Therefore, it has great economic and social value for the status monitoring and fault diagnosis of wind turbines.
In this paper, the empirical mode decomposition and improved bispectrum vibration signal analysis method are combined with the monitoring method of the inspection, which is introduced into the fault diagnosis process of the fan gearbox, and combined with the production example for fan fault diagnosis.

Signal processing methods
The vibration signal of wind turbine has strong nonstationarity and nonlinear (caused by coupling of vibration signals during meshing of multiple pairs of gears).This brings great difficulty to the fault diagnosis of the gearbox of the wind turbine. The vibration signal of wind turbine is very complicated so it is necessary to use powerful signal analysis tools to extract useful information.

Empirical mode decomposition (EMD)
The EMD algorithm can decompose the input signal into several narrowband components (IMFs) of different frequencies, and the result of the decomposition is several IMFs and one residual signal:

Residual signals
The eigenmode function should satisfy two conditions: ①The difference between the sum of the maxima and minima points contained in the eigenmode function and the number of zero crossings is not more than 1; ②The sum of the upper and lower envelopes of all points is zero among the eigenmode function signals.
The empirical mode decomposition algorithm can be divided into the following 2 steps: ① Solving the IMF component of   is not greater than the pre-calculation set value or the residual quantity () n rt of the nthorder eigenmode function is When the monotonic function is completed, the entire empirical mode decomposition process ends.
Bispectrum is defined as The bispectrum is often normalized to obtain the bis-coherent spectrum.

Improved double spectrum.
In order to solve the deficiency of bispectral analysis of amplitude modulation signal Stack J R proposed an improved bispectrum. The definition is as follows: The normalized form is:

Diagnostic example
First, the first test and analysis are performed as reference data.   It can be seen from this that the vibration signal of the fan is very complex and it takes powerful signal analysis tools to extract useful information. Figure 3 is the result of empirical mode decomposition (EMD) of the vibration signal. The left side of the diagram is the intrinsic modal function component of each order column and the right side is the corresponding power spectrum. The uppermost edge is the IMF component which is decomposed from the original sequence with the smallest amplitude and the highest frequency. The amplitude of each IMF component increases gradually and the frequency decreases gradually until the very low frequency component. By observing the decomposition process of EMD we can find that EMD can be regarded as an adaptive filter bank. Each IMF component is a "feature component" contained in the original signal. The bandwidth is determined by the characteristics of the signal itself. EMD is a good signal preprocessing method because it is carried out completely in time domain.

Comparative Analysis of Vibration Detection Data of B07 Unit
The vibration amplitude of the ring gear of B07 has increased significantly. Under the same conditions (wind speed 6.7m/s, wind wheel speed 11.8r/min, high speed shaft speed 1230r/min), the effective value of vibration (acceleration) increased from the last 0.724m/s 2 to 1.01m/s 2 . Perform

Conclusion
In this paper, the method of signal analysis by decomposing empirical mode and improving bispectrum is studied combining with the inspiration mode in order to fault diagnosis and condition monitoring for Gearbox of Wind Turbine. It has been proved that the work condition of wind turbine unit can be monitored and fault diagnosis can be made in the way of inspection. Ensure long-term safety and full load operation of wind turbine.
It is proved that the empirical mode decomposition and improved bispectral signal processing method are suitable for the fault diagnosis of wind turbine gearbox. The empirical mode decomposition (EMD) and improved bispectrum are applied to the complicated signal fault diagnosis of the gearbox of wind turbine which can effectively and accurately separate the fault characteristic information.