滚动轴承故障诊断的最优加权路图GFT冲击提取方法

Impulse extraction by optimal weighted path GFT method for roller bearing fault diagnosis

滚动轴承发生局部故障时将产生由共振频率调制的周期瞬态冲击,有效提取冲击特征是诊断滚动轴承故障的关键。图信号处理方法(Graph Signal Processing,GSP)是基于图谱理论发展起来的新研究领域,将振动信号转换为图信号进行分析,能有效揭示振动信号特性。对高斯函数加权下的路图拉普拉斯矩阵进行特征分解,发现代数连通度(Algebraic Connectivity)以内的特征向量存在明显的冲击,因此提出利用代数连通度以内的特征向量结合逆图傅里叶变换(Graph Fourier Transform,GFT)重构故障信号中冲击分量的方法。高斯加权函数中的热核宽度决定冲击特征向量的分布,直接影响重构结果,为解决热核宽度的选择问题,提出结合粒子群算法(Particle Swarm Optimization,PSO)确定最优热核宽度;然后利用最优特征向量组重构冲击信号,并进行包络解调;最后实现滚动轴承故障的有效诊断。算法仿真和应用实例表明,基于最优加权的路图GFT方法能有效地重构滚动轴承故障冲击特征,诊断故障类型。

Local defects in bearings give rise to periodic transient impulses modulated by response frequency.Therefore,effective extraction of impulse signal is the key to diagnosing rolling bearing fault.Graph signal processing is a new research domain based on graph spectrum theory.The analysis of graph signal transformed from vibration signal can effectively reveal vibration signal characteristics.After the eigen-decomposition of path graph Laplacian matrix,the eigenvectors within the algebraic connectivity have impulse feature.In this paper,an impulse component reconstruction method by using eigenvectors within the algebraic connectivity and the inverse graph Fourier transform is proposed and applied to the fault diagnosis of roller bearings.The thermonuclear width in the Gaussian function determines the distribution of the impulse eigenvectors and directly affects the reconstruction results,thus,the particle swarm optimization algorithm is used to get the optimal thermonuclear width.The impulse component is then reconstructed by selected eigenvector group and the faults of rolling bearings can be diagnosed by envelope demodulation.The simulation and application examples have proved that the proposed method is effective in reconstructing the impulse feature of rolling bearing and diagnosing the faults type.