基于SQSWOMP算法的轴承实时故障诊断研究  

Bearing real-time fault diagnosis based on SQSWOMP algorithm

作　　者：钱子君 吴波[1] 于剑峰 王文瑞 王振明 袁晓兵[3] QIAN Zi-jun;WU Bo;YU Jian-feng;WANG Wen-rui;WANG Zhen-ming;YUAN Xiao-bing(Shanghai Advanced Research Institute,Chinese Academy of Sciences,Shanghai 200120,China;University of Chinese Academy of Sciences,Beijing 100089,China;Shanghai Institute of Microsystem and Information Technology,Chinese Academy of Sciences,Shanghai 200233,China)

机构地区：[1]中国科学院上海高等研究院,上海200120 [2]中国科学院大学,北京100089 [3]中国科学院上海微系统与信息技术研究所,上海200233

出　　处：《机电工程》2022年第10期1333-1344,1355,共13页Journal of Mechanical & Electrical Engineering

基　　金：上海市自然科学基金资助项目(19ZR1463800);中国科学院上海微系统与信息技术研究所微系统技术重点实验室基金资助项目(6142804190206)。

摘　　要：针对轴承故障诊断方法存在数据存储、传输、处理效率低,以及模型参数量大导致资源浪费、故障诊断精度低等问题,在基于改进的分段弱正交匹配追踪(SWOMP)算法的基础上,针对SWOMP算法存在重构精度低的问题,在SWOMP算法中引入S形函数与对原子支撑集的二次选择过程,提出了一种基于S形二次分段弱正交匹配追踪(SQSWOMP)算法的轴承实时故障诊断方法。首先,采用滑动移窗,将轴承历史数据分割成矩阵,得到了其初始字典,用改进的K-SVD算法建立了过完备字典;然后,将SQSWOMP算法与训练好的过完备字典结合,重构了轴承的实时振动信号;最后,利用字典可以有效对特定状态信号进行稀疏分解,而对其他状态信号不能稀疏分解的特点,对重构信号进行了分类。研究结果表明:相比于其他匹配追踪算法,SQSWOMP算法在重建精度和效率方面具有优势,解决了SWOMP算法重建精度低的问题;在不需要消耗大量存储、传输资源的情况下,采用SQSWOMP算法可将正常、内圈、外圈、滚动体故障诊断准确率提升2.17%、23.46%、17.30%、18.38%,且其准确率也优于其他方法,为轴承的实时故障诊断提供了便利。Aiming at the problems in the bearing fault diagnosis method,such as low efficiency of data storage,transmission and processing,and large number of model parameters leading to waste of resources and low diagnostic accuracy.Based on the improved stagewise weakly orthogonal matching pursuit(SWOMP)algorithm,in view of the problem of low reconstruction accuracy in the SWOMP algorithm,a sigmoid function and a secondary selection process for the atomic support set were introduced into the SWOMP algorithm.A real-time fault diagnosis method for bearings based on S-shaped quadratic stagewise weak orthogonal matching pursuit(SQSWOMP)algorithm was proposed.Firstly,the initial dictionary was obtained by partitioning the bearing history data into matrices with sliding shift windows,while the overcomplete dictionary was established with the improved K-SVD algorithm.Then,SQSWOMP algorithm was combined with the trained overcomplete dictionary to reconstruct the real-time vibration signals.Finally,the reconstructed signals were sparsely decomposed by using the feature that the dictionary could effectively decompose the specific state signals,but not the other state signals.The results indicate that the reconstruction accuracy and efficiency of SQSWOMP algorithm are better than other matching and tracking algorithms,which can solve the problem of low reconstruction accuracy of SWOMP algorithm.The accuracy of normal,inner ring,outer ring,and rolling body fault diagnosis are improved by 2.17%,23.46%,17.30%,and 18.38%without consuming a lot of storage and transmission resources.The accuracy is better than other methods,which can facilitate real-time fault diagnosis.

关 键 词：分段弱正交匹配追踪算法 S形二次分段弱正交匹配追踪算法 压缩感知理论 信号重构 稀疏表示 改进贪婪算法 

分 类 号：TH133.33[机械工程—机械制造及自动化]