基于压缩感知的缺失机械振动信号重构新方法  

Novel method for missing mechanical vibration signal reconstruction based on compressed sensing

作　　者：郭俊锋[1] 胡婧怡 王智明[1] GUO Junfeng;HU Jingyi;WANG Zhiming(School of Mechanical and Electronic Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学机电工程学院,兰州730050

出　　处：《振动与冲击》2024年第10期197-204,共8页Journal of Vibration and Shock

基　　金：甘肃省高校科研创新平台重大培育项目(2024CXPT-04)。

摘　　要：针对工业机械设备实时监测中不可控因素导致的振动信号数据缺失问题,提出一种基于自适应二次临近项交替方向乘子算法(adaptive quadratic proximity-alternating direction method of multipliers, AQ-ADMM)的压缩感知缺失信号重构方法。AQ-ADMM算法在经典交替方向乘子算法算法迭代过程中添加二次临近项,且能够自适应选取惩罚参数。首先在数据中心建立信号参考数据库用于构造初始字典,然后将K-奇异值分解(K-singular value decomposition, K-SVD)字典学习算法和AQ-ADMM算法结合重构缺失信号。对仿真信号和两种真实轴承信号数据集添加高斯白噪声后作为样本,试验结果表明当信号压缩率在50%~70%时,所提方法性能指标明显优于其它传统方法,在重构信号的同时实现了对含缺失数据机械振动信号的快速精确修复。In order to address the issue of missing vibration signal data in real-time monitoring of industrial machinery due to uncontrollable factors,a compressed sensing missing signal reconstruction method based on the adaptive quadratic proximity-alternating direction method of multipliers(AQ-ADMM)was proposed.In the AQADMM algorithm,a quadratic proximity term was introduced into the classic alternating direction method of multipliers iterative process and penalty parameters were adaptively selected.First,a signal reference database was established at the data center for creating an initial dictionary.Then,the missing signals were repaired using a reconstruction method based on the K-singular value decomposition(K-SVD)dictionary learning algorithm and the AQ-ADMM.Gaussian white noise was added to the simulated signals and two real bearing signal datasets to serve as samples.The experimental results demonstrate that the proposed method presents significantly better performance indices than other traditional methods when the signal compression ratio ranges from 50%to 70%.It achieves fast and accurate recovery of missing data signals while reconstructing the signals.

关 键 词：压缩感知 缺失信号 自适应二次临近项交替方向乘子算法(AQ-ADMM) K-奇异值分解(K-SVD) 正交匹配追踪 

分 类 号：TN911.7[电子电信—通信与信息系统] TH113.1[电子电信—信息与通信工程]