Research on Multiwavelet Preprocessing Methods for One Dimensional Signal Analysis
XU Yong-gang · Machinery Design and Manufacture · 2012
Due to owing the characteristics of symmetry,orthogonality,short support and high order vanish moments,etc.simultaneously,multiwavelet is much more advantageous in impulse feature extraction than traditional wavelet transform.However,the multiwavelet is a multidimensional input and output system,i.e.the low-pass and high-pass filter banks are both composed of r×r matrices.As a result,the problem needed to be solved first is the preprocessing on the rough data when the analyzed signal is one dimensional vibration signal.It will deal with two kinds of signal,i.e.noischir simulation signal and vibration data of some rolling element bearing,to study the influence of the preprocessing on the signal,in which,the reconstruction error and energy ratio of low frequency coefficients after multiwavelet decomposition will be regarded as the evaluation indices.At last,the strategy for selecting rational preprocessing methods will be provided.