基于张量奇异谱分解的机械故障特征提取方法研究

摘要

提出了将张量奇异谱分解运用于机械故障诊断特征提取,张量奇异谱分解(TSSA)是传统奇异谱分解(SVD)的扩展.由于传统奇异谱分解在处理非平稳、非线性的信号效果不理想,故障特征不明显.因此将传统的奇异谱分解延伸到三阶张量分解中,从而将一维的时间序列转换成为三阶的张量,然后运用标准(PARAFAC)张量分解,标准(PARAFAC)张量分解是把秩为R的张量分解为R个秩-1的张量的和,分解出原始张量的因子矩阵和权重,并重构回一维信号进而对信号的时域和频域做出分析.为了证明方法的有效性,将该方法应用于轴承故障信号的特征提取中,分别运用了仿真和实测信号做了分析,结果表明该方法不仅能有效地抑制噪声,明显地提取轴承故障信号特征,而且其效果要优于传统的奇异谱分解方法,具有一定的工程实践价值.%Tensor singular spectrum decomposition (TSSA) is applied to mechanical fault,which is the extension of the traditional singular spectrum decomposition (SVD).The traditional singular spectrum decomposition is not good at process the non-stationary and nonlinear signal,thus the fault feature is not obvious.In order to overcome the weakness,singular spectrum decomposition is extended to the three order tensor decomposition.After the one-dimensional time series is converted into a three order tensor,the factor matrix and the weight of the original tensor is decomposed by using standard (PARAFAC) tensor.The standard (PARAFAC) tensor decompositiondecomposes the rank of Rtensor into the sum of rank-1 tensor.Then we reconstruct the one-dimensional signal and analyze its time and frequeney domain.In order to prove the validity of the proposed method,this method is applied to the feature extraction of bearing fault signal.The result shows that the proposed method can effectively reduce the noise and extract the information of bearing fault clearly.Therefore,it has a certain value in the engineering practice.