基于FSDPC_Otsu算法的滚动轴承故障研究

Rolling Bearing Fault Diagnosis Based on Clustering by Fast Search and Find of Density Peaks Combined Otsu Method

针对振动源数未知且观测信号小于振动源数量的欠定盲源问题,提出一种改进快速寻找密度峰值聚类(FSDPC)的方法。首先将混合信号投影到多维空间上并计算每点的密度值,在此基础上利用最大类间方差法(Otsu)对点密度进行阈值分割,去除干扰点对聚类准确率的影响;然后根据数据的密度峰值确定聚类中心,估计混合矩阵;最后通过L_(1)范数最小化对混合信号进行分离并进行包络谱分析,实现轴承故障诊断。FSDPC_(O)tsu方法可在源数和聚类中心初值未知的条件下估计混合矩阵,且保证混合矩阵精度。实验结果表明,应用FSDPC_(O)tsu方法的稀疏成分分析能够对轴承多故障信号进行欠定盲分离,进而实现故障识别与诊断。

To solve the problem of underdetermined blind separation,caused by unknown vibration sources and smaller number of observation signals,an improved method of clustering by fast search and find of density peaks( FSDPC)is proposed.Initially,the mixed signal is projected to the multi-dimensional space and then calculate each point density value,using the Otsu method following for density threshold segmentation so as to remove the influence of interference on the accuracy of clustering,and then determine the cluster center according to the data density peaks,to estimate the mixing matrix;finally utilizing L_(1) norm minimization to separate mixed signals,and the envelope spectrum analysis is carried to realize fault diagnosis.The FSDPC_(O)tsu method can estimate the mixed matrix under the condition of the unknown source number and the initial value of the cluster center,at the same time,the accuracy of the mixed matrix can be guaranteed.The experimental results show that the sparse component analysis of the FSDPC_(O)tsu can separate the multiple fault signals of the bearing and realize the fault diagnosis.