基于VMD-PARAFAC的轴承故障欠定盲源分离

针对传统的轴承故障欠定盲源分离方法需要施加约束的问题,提出了一种基于变分模态分解(variational mode decomposition,VMD)和平行因子(parallel factor,PARAFAC)分析的欠定盲源分离方法.利用VMD算法将振动信号分解为多个带限本征模态函数(band-limited intrinsic mode functions,BLIM Fs),将这些BLIM Fs构造成三阶张量作为PARAFAC模型的输入,利用三线性交替最小二乘算法对模型分解,从而在宽松条件下实现复合故障信号的分离.仿真和实验结果表明,提出的方法是有效的,与传统的故障盲源分离方法比较,提出的方法在多故障盲源分离中更具有适应性和实用性.

Underdetermined blind source separation of bearing faults based on VMD-PARAFAC

Aiming at the problem that traditional underdetermined blind source separation methods for bearing faults need to impose some constraints, a new underdetermined blind source separation method based on variable mode decomposition(VMD)and parallel factor analysis(PARAFAC)was proposed. VMD algorithm was used to decompose the vibration signal into band limited intrinsic mode functions(BLIMFs), and then these BLIMFs were constructed into three-order tensor as the input to PARAFAC model. The model was decomposed by trilinear alternating least square algorithm, so that the compound fault signal could be separated in loose conditions. The simulation and experimental results show that the as-proposed method is very effective. Compared with the traditional blind source separation method, the as-proposed method is more adaptive and practical in the blind source separation of multi-fault.