基于稀疏分解的振动信号数据压缩算法

发布时间：2018-01-13 17:37

本文关键词：基于稀疏分解的振动信号数据压缩算法　出处：《仪器仪表学报》2016年11期 　论文类型：期刊论文

【摘要】：针对齿轮传动装置在状态监测与故障诊断过程中面临的大量振动信号传输困难问题,提出利用K-SVD算法进行信号的稀疏分解,进而完成对大量振动数据的压缩。传统K-SVD算法在字典更新过程中对时间的消耗量较大,特别是在大量振动数据压缩过程中,对数据压缩效率较低,为此提出一种K-SVD字典更新的改进算法。改进算法从单次迭代过程中参与更新的字典原子列数出发,每次奇异值分解后对多列字典原子同时进行赋值,从而减少单次迭代计算量。根据不同原子列数在稀疏分解过程中的迭代收敛次数、时间消耗与重构峰值信噪比,以此确定最佳的字典更新列数。实验结果表明:传统K-SVD算法对振动信号的数据压缩效率较低,改进算法能够在保证信号压缩比与重构效果的前提下,有效缩短训练字典的时间消耗。
[Abstract]:According to the gear face in the condition monitoring and fault diagnosis of large vibration signal transmission problems, proposed sparse signal using K-SVD algorithm and then complete the decomposition, compression of large vibration data. The consumption of time in the dictionary update process between the traditional K-SVD algorithm is large, especially in a large number of vibration data compression process the data compression, the efficiency is low, so this paper proposes an improved K-SVD algorithm. The improved algorithm to update the dictionary from a single iteration in the update of the dictionary of atomic number of columns, each singular value decomposition of multiple dictionary atoms at the same time of assignment, so as to reduce the amount of calculation of single iteration. According to the different number of columns in the atomic sparse in the process of decomposition of the iteration number, time consumption and reconstruction of peak signal-to-noise ratio, in order to determine the best dictionary update column number. The experimental results show that the traditional K-SVD The data compression efficiency of algorithm for vibration signal is low, and the improved algorithm can effectively shorten the time consumption of training dictionary under the premise of ensuring signal compression ratio and reconstruction effect.

【作者单位】：军械工程学院;
【基金】：国家自然基金(E51305454)项目资助
【分类号】：TH132.41
【正文快照】： 1引言大型机电设备中,齿轮传动为主要传动方式,其工作状态的好坏,直接影响到整个机电设备的运行情况。目前,针对齿轮传动的状态监测与故障诊断主要以振动信号为依据,采用分布式的状态监测系统[1-2]。但是在对齿轮传动装置实时监测过程中,振动信号数据量巨大,数据结构复杂,给数

【相似文献】