基于盲源分离和AR谱估计的旋转机械故障诊断方法

准确的AR模型能够较好地揭示信号中蕴含的状态特征变化的信息,然而,AR模型对系统的状态变化十分敏感,多个动态变化的源信号的耦合必然会影响其估计结果。基于此,提出了一种基于盲源分离和AR谱估计的旋转机械故障诊断方法。首先,利用盲源分离的方法从混合观测信号中恢复各机械振动源信号;然后,将非平稳性的故障信号通过经验模态分解得到各本征模态函数;最后,对经验模态分解得到的平稳的本征模态函数进行AR谱估计,提取振动信号的故障特征信息。通过仿真研究和实验分析验证了该方法在旋转机械故障诊断中的有效性和可行性。

The Fault Diagnosis for Rotating Machinery Based on BSS and AR Spectrum Estimation

The accurate AR model can reveal the changing state characteristics inherent in the signal, however the AR model is sensitive to the changes in the state of the system, and the multiple of dynamic source signal coupling is bound to affect the estimated results. The method of blind source separation is reconstruct mechanical vibration source signals. Then the non-stationary fault signal is decomposed into several stationary signals which suit to establish AR model. Finally, the AR model of stationary intrinsic mode function is constructed to extract the characteristics of fault vibration signal. The results of simulation and experiment are presented to verify the theory analysis.