基于改进MP的稀疏表示快速算法及其滚动轴承故障特征提取应用

摘要

At constant rotating speed,localized faults of rolling bearings in rotating machines may lead to periodic impacts and thus cause periodic transient vibration signals.The transient feature extraction always is a crucial problem for localized fault detection.The sparse decomposition based on matching pursuit (MP)is an adaptive sparse representation for signals.It is one of the effective methods for weak feature extraction.Here,aiming at the selecting and constructing problems of the over complete dictionary for sparse representation of rolling bearing fault vibration signals,the matching correlation filtering method was used to construct Laplace wavelet atoms used to match the impact waveform and construct the over-complete dictionary.Aiming at the large computational cost and low efficiency problems of the basic matching tracking algorithm,combined with FFT fast operation characteristics,the correlation algorithm was used to replace the basic matching pursuit algorithm for inner product computation to improve the computational efficiency.Simulation and test analysis results of rolling bearings with faults showed that the proposed algorithm can extract their fault features accurately and efficiently.%在恒定转速情况下，旋转机械中滚动轴承的局部故障往往导致周期性冲击，从而产生周期性瞬态振动信号。对局部故障的瞬态特征提取一直是故障检测的关键问题。基于匹配追踪（Matching Pursuit，MP）算法的稀疏分解是一种信号自适应分解算法，是强噪声背景下微弱特征提取的有效方法之一。针对滚动轴承故障振动信号稀疏表示过完备字典的选择与构造问题，基于相关滤波法优选与冲击波形匹配的 Laplace 小波原子构造稀疏表示中的过完备字典；针对基本匹配追踪算法计算量大、效率低的问题，结合 FFT 快速运算特性，通过互相关运算替换基本匹配追踪算法中的内积运算，研究基于改进 MP 的稀疏表示快速算法，进而提高计算效率。仿真与滚动轴承故障实验分析结果表明该算法能准确的提取滚动轴承故障特征且计算效率高。