| As an important connecting part of the wind turbine to transform wind energy into electric energy, the wind turbine blade has attracted more and more attention of the industry.How to detect the blade’s health state is an important part in the blade condition monitoring. Based on the complex vibration signal and the difficulty to recognize the crack signal,this paper has proposed a method of bispectrum analysis of intrinsic mode function to extract the crack signal. The main contents are the following several aspects:Firstly, the blade vibration signal characteristics has been studied through the mathematical model of the cantilever beam structure of blade.Secondly, the method of kernel principal component analysis(KPCA)be adopted to reduce the White noise and interference signals in one dimensional time domain signal,for white noise and interference signal will impact on bispectrum analysis. Mutual information method and Cao’s method be used for one dimensional time series phase space reconstruction, the principal component of the phase space be extracted through the kernel matrix, one-dimensional signal noise finally be reduced by the inverse transformation of phase space reconstruction.Thirdly, the characteristics of intrinsic mode function(IMF) and bispectrum has been studied, in view of the nonlinear and coupling blade vibration signal, the method of bispectrum analysis of the reconstruction of intrinsic mode function be put forward to identify the crack signals. The bispectrum’s inhibitory effect on the gaussian white noise be studied, and it’s recognition effect on Impact signal and coupling signal in non-stationary signals be researched too.Finally, the blade vibration experiment is simulated relying on wind turbine blade fatigue test platform which locate in Shanghai zhiyuan co.LTD. The analysis of the experimental data show that the blade cracks will produce impact component and coupling component, these features will be extracted well by bispectrum analysis. |
PDF Full Text Request