Mechanical impact faults extraction method based on nonlinear manifold learning for continuous wavelet coefficients
LI Mao-lin1; LIANG Lin1,2; WANG Sun-an1; ZHUANG Jian1
1. School of Mechanical Engineering，Xi’an Jiaotong University，Xi’an 710049，China；
2. The State Key Laboratory for Manufacturing Systems Engineering，Xi’an Jiaotong University，Xi’an 710049，China
Abstract：To acquire the impact component aroused by mechanical fault, a novel feature extraction method based on nonlinear manifold learning for continuous wavelet coefficients is put forward. Firstly, the wavelet entropy method is adopted to optimize the Morlet wavelet shape factor in order to match with the impact components to obtain the optimal continuous wavelet coefficients. Secondly, the nonlinear manifold learning algorithm named local tangent space alignment is used to the reduction analysis of the optimal wavelet coefficients matrix, and according to the principle of the maximum kurtosis index, the low-dimensional embedded vectors which introduced to reflect the impact of the failure are extracted from the global coordinates feature matrix. Finally, the simulation and industrial applications show that this approach, compared with the singular value decomposition, is effective to extract not only the weak periodic impacts with the greater kurtosis in time waveform, but also the fault feature frequency in frequency spectrum.
2012, Vol. 31 
