Extraction of fault feature in gear system based on convolution type of wavelet packet transform and singular value decomposition

A novel approach to extract fault feature parameters is put forward. First, the time signal is transformed to time-frequency signals which keep same length as that of the original signal by using the convolution type of wavelet packet transformation.Second, considering time-frequency signals as the matrix reflect feature of system, singular value decomposition (SVD) is used to convert the multi-dimension time-frequency matrix to one dimension feature vector. Last, effective feature extraction is achieved. At the same time, an improved singular value decomposition method (ISVD) is employed. ISVD overcomes shortcomings of classical method which canpsilat determine corresponding relation of singular value with row vectors of input matrix, and ensures precise of feature information.