在重构相空间重构吸引子的基础上,提出一种基于奇异谱分析的设备故障诊断方法。基于奇异谱理论介绍了该方法的基本原理,并给出该方法的实现步骤。实验研究表明,该方法可以显著降低噪声信号,有效地提高信噪比,并且能保留与噪声信号的频谱相混叠的系统信号,突出故障的信息特征,可大大提高设备故障诊断的准确性,它同时还具有良好的瞬态信息提取能力。
A fault diagnosis method based on singular spectrum analysis is described based on the reconstructed attractor in reconstructed phase space. The basic principle and of this method realization steps are introduced according to the singular spectrum theory. Experiment shows that this method can reduce the noise and improve SNR effectively. It can remain the system signal whose frequency spectrum is mixed with the noise. And it can extrude the fault characteristic. So the veracity of mechanical device fault diagnosis can be improved greatly. Furthermore, this method also has the ability to distill transient signals information.