优化奇异谱分解方法在轴承故障诊断中的应用

Application of Optimized Singular Spectrum Decomposition Method in Fault Diagnosis for Bearings

奇异谱分解在处理强噪声信号时获得的模态分量可能包含期待频段之外的信息,会造成严重的模态混叠现象并影响分析效果,深入研究发现造成上述现象的原因是迭代过程中轨迹矩阵的嵌入维数设定不合理。在大量数据分析的基础上提出了一种优化的奇异谱分解方法(OSSD),以迭代过程中划分的频段及重构分量时特征向量的选择为依据确定新的参数并设定嵌入维数,不仅可以使构造的轨迹矩阵更加合理,还可以使分量的重构更加准确。仿真及试验分析表明,该方法可以有效抑制模态混叠现象,减少分解所得分量在频域上的能量泄漏,准确提取滚动轴承振动信号中的故障特征。

The modal components obtained by Singular Spectrum Decomposition(SSD)may contain information beyond expected frequency band when processing strong noise signals,which will cause the serious mode aliasing phenomenon and affect the analysis effect.In-depth study finds that the above phenomenon is caused by unreasonable setting of embedding dimension of trajectory matrix during iteration process.Based on a large number of data analysis,the Optimized SSD(OSSD)method is proposed.The proposed method determines the new parameters and sets the embedding dimension based on frequency band divided in iteration process and selection of eigenvectors during reconstruction of components,which can not only make the constructed trajectory matrix more reasonable,but also make the reconstruction of components more accurate.The simulation and experimental analysis show that the method can effectively suppress the mode aliasing phenomenon,reduce the energy leakage of decomposed components in frequency domain,and accurately extract fault features from rolling bearing vibration signals.