基于FastICA的旋转机械故障特征盲源分离方法研究

在快速独立分量分析基础上,提出了一种基于改进牛顿迭代法的独立分量分析算法,利用所建立的方法进行振动信号盲源分离的数值仿真,并对具有故障的实际转子进行多传感器信号采集并进行盲分离。仿真实验及实验室实测振动信号的分离结果证明该方法是有效的,可作为旋转机械故障诊断的信号预处理方法。

Blind Source Separation of Rotating Machine Faults Based on FastICA

A new FastICA algorithm based on Newton iteration method is proposed.Based on the method developed,the numerical simulation of vibration signal blind source separation was performed.And this blind source separation method was applied to multi-sensor signals of a real rotor faults.The computer simulation and experimental verification show that the algorithm can well restore source signals from their mixtures.This method can be seen as a pre-processing step that improves the diagnosis for rotating machine.