针对广义正交匹配追踪算法(GOMP)在进行滚动轴承振动信号压缩感知重构的迭代过程中无法剔除错误原子,重构效果较差的问题,提出了基于麻雀搜索算法-回溯广义正交匹配追踪(SSA-BGOMP)的轴承振动信号压缩重构方法,在GOMP的基础上引入具有自适应特性的改进回溯机制,通过麻雀搜索算法自动设置阈值,对支撑集原子进行二次回溯筛选,从而降低错误原子选入支撑集的概率,提升算法的抗噪性和重构效果.仿真信号以及CWRU,XJTU-SY轴承故障数据集的试验结果表明:在DCT和K-SVD字典上,SSA-BGOMP比GOMP的相对误差分别降低2%~12%与3%~13%,有效改善了滚动轴承振动信号的压缩重构效果.
Aimed at the problem that the generalized orthogonal matching pursuit(GOMP)cannot eliminate the wrong atoms during iterative process of compressed sensing reconstruction of vibration signals for rolling bearings,and the reconstruction effect is poor,a compressed reconstruction method for vibration signals of the bearings is proposed based on sparrow search algorithm-back GOMP(SSA-BGOMP).An improved backtracking mechanism with adaptive characteristics is introduced on the basis of GOMP,and the threshold is automatically set by sparrow search algorithm to perform secondary backtracking screening on atoms of support set,thus reducing the probability of wrong atoms being selected into support set and improving the noise resistance and reconstruction effect of the algorithm.The experimental results of simulation signals and CWRU and XJTU-SY bearing fault datasets show that the relative errors of SSA-BGOMP are reduced by 2%~12%and 3%~13%respectively compared with GOMP on DCT and K-SVD dictionaries,which effectively improves the compressed reconstruction effect of vibration signals for rolling bearings.