基于VMD和BiCS的机械状态监测数据修复方法  

Data restoration method of bearing vibration signal based on VMD and BiCS

作　　者：李帅永 傅圣豪 梅琳 高光勇 LI Shuaiyong;FU Shenghao;MEI Lin;GAO Guangyong(SWS Hemodialysis Care Co.,Ltd,Chongqing 401121,China;Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Chongqing Special Equipment Inspection and Research Institute,Chongqing 401121,China)

机构地区：[1]重庆山外山血液净化技术股份有限公司,重庆401121 [2]重庆邮电大学,重庆400065 [3]重庆市特种设备检测研究院,重庆401121

出　　处：《自动化与仪器仪表》2023年第3期55-62,共8页Automation & Instrumentation

基　　金：国家自然科学基金资助项目(61703066,51705056);重庆市博士后研究项目特别资助(2021XM3026);国家市场监管重点实验室(西部复杂环境机电设备安全)开放基金(CQTJ-XBJD-KFKT202202)。

摘　　要：工业“大数据”时代的到来为机械装备健康监测带来了新机遇。然而,由于运行环境异常、人为因素干扰以及采集设备故障等,机械装备健康监测大数据中往往混杂大量与健康状态无关的异常值或缺失值数据,从而造成数据质量下降。为保障数据质量,提高振动信号恢复效果,提出一种基于变分模态分解和双向压缩感知(Variational Mode Decomposition, VMD;Bidirectional Compressed sensing, BiCS;VMD-BiCS)的振动信号重构方法。首先通过变分模态分解对采样数据进行去噪处理,其次在压缩感知框架下构造观测矩阵、稀疏表示字典矩阵,然后基于子空间追踪(Subspace Pursuit, SP)算法从两个方向重构出稀疏表示向量。在此基础上利用离散余弦变换构造稀疏矩阵,利用压缩感知原理重构双向信号,最后对重构信号加权得到最终重构信号,实现对残损振动信号的数据恢复。分别采用仿真信号和西储大学公开轴承数据进行修复实验,将该方法用于对压缩感知传统重构算法进行改进,发现所提方法在时域指标均方根误差上均优于传统重构算法。并从修复效果角度验证发现该方法成功还原了外圈故障信号的故障特征频率,达到了修复的目的。The arrival of industrial “big data” era has brought new opportunities for mechanical equipment health monitoring. However, due to abnormal operating environment, human factor interference and acquisition equipment failure, a large number of abnormal value or missing value data irrelevant to health status are often mixed in the big data of mechanical equipment health monitoring, resulting in the decline of data quality. In order to ensure the data quality and improve the effect of vibration signal repair, a vibration signal reconstruction method based on variational mode decomposition and bidirectional compressed sensing (VMD;BiCS;VMD BiCS) is proposed in this paper. Firstly, the sampled data are degraded and denoised by variational modal decomposition. Secondly, the observation matrix and sparse representation dictionary matrix are constructed under the compressed sensing framework, and then the sparse representation vector is reconstructed from two directions based on subspace pursuit (SP) algorithm. On this basis, the discrete cosine transform is used to construct the sparse matrix, and the compressed sensing principle is used to reconstruct the bidirectional signal. Finally, the reconstructed signal is weighted to obtain the final reconstructed signal, so as to realize the data recovery of the damaged vibration signal. The simulation signal and the public bearing data of Western Reserve University are used to improve the traditional reconstruction algorithm of compressed sensing. It is found that the proposed method is better than the traditional reconstruction algorithm in the root mean square error of time-domain index. From the perspective of repair effect, it is verified that this method successfully restores the fault characteristic frequency of outer ring fault signal, and achieves the purpose of repair.

关 键 词：变分模态分解 压缩感知 振动信号 数据恢复 

分 类 号：TN911.7[电子电信—通信与信息系统] TH86[电子电信—信息与通信工程]