Compressed sensing reconstruction for axial piston pump bearing vibration signals based on adaptive sparse dictionary model

Xiao Chaoang,Ren Yan,Tang Hesheng

doi:10.1177/0020294019898725

Abstract

Aiming at the mechanical equipment in the fault diagnosis process, the traditional Shannon–Nyquist sampling theorem is used for data collection, which faces main problems of storage, transmission, and processing of mechanical vibration signals. This paper presents a novel method of compressed sensing reconstruction for axial piston pump bearing vibration signals based on the adaptive sparse dictionary model. First, vibration signals were divided into blocks, and an energy sequence was produced in accordance with the energy of each signal block. Second, the energy sequence of each signal block was classified by the quantum particle swarm optimization algorithm. Finally, the reconstruction of machinery vibration signals was carried out using the K-SVD dictionary algorithm. The average relative error of the reconstructed signal obtained by the proposed algorithm is 4.25%, and the reconstruction time decreases by 43.6% when the compression ratio is 1.6.

Highlights

Swashplate-type axial piston pumps are one of the power-dense types of hydraulic units and are widely used in industrial fields.[1]
To improve the reconstruction effect of vibration signal, a new method of compressed sensing reconstruction for axial piston pump bearing vibration signals based on adaptive sparse dictionary model is proposed in this study
Vibration signals are divided into blocks, and an energy sequence was produced in accordance with the energy of each signal block

Summary

Introduction

Swashplate-type axial piston pumps are one of the power-dense types of hydraulic units and are widely used in industrial fields.[1]. The nonsparse signals can convert to sparse signals based on transformations, for example, Fourier transform (FT), wavelet transform (WT),[29] and discrete cosine transform (DCT).[30] the signal can be reconstructed by appropriate recovery algorithms that can solve unconstrained optimization problems.[31,32,33] Wang et al.[34] proposed a new sparsity empirical WT method to detect bearing faults. The proposed framework mainly involves three processes as follows: (1) vibration signal compression acquisition, (2) adaptive sparse dictionary model, and (3) signal reconstruction. The product FcÀ1 satisfies RIP with a high probability if F satisfies RIP, and c is orthogonal

Design of adaptive sparse dictionary

Findings

Conclusion