Vibration data reconstruction method based on variable step size stepwise adaptive matching pursuit

WANG Peng-fei, SHENG Bu-yun
(School of Mechanical and Electrical Engineering, Wuhan University of Technology, Wuhan 430070, China)

Abstract: Aiming at the problem of huge storage and transmission pressure in processing multidimensional massive vibration data, the theory of compressed sensing was studied, and based on the optimal measurement matrix, a variable step size stepwise adaptive matching pursuit (VSSStAMP) was proposed. It was combined with matching pursuit, variable step size iteration, and sparsity adaptive, and signal reconstruction accuracy was controlled by double thresholds and variable step size. The shortage was covered that the traditional reconstruction algorithms need to obtain the signal sparsity in advance, and the sparsity adaptive algorithm (SAMP) reconstruction result was greatly affected by the step size, to realize the reconstruction of the vibration signal adaptively. The results indicate that compared with the traditional reconstruction algorithm and SAMP algorithm, the VSSStAMP algorithm improves in terms of mean square error and running time, and effectively improves the accuracy and efficiency of vibration signal reconstruction.
Key words: variable step size stepwise adaptive matching pursuit（VSSStAMP）; mechanical vibration; vibration signal sampling; measurement matrix; reconstruction algorithm