转子振动故障信号的盲分离

介绍了盲源信号分离基本原理。基于最小化互信息原理,建立了用独立分量分析方法估计分离矩阵的加速梯度法,并给出了实现步骤。利用所建立的方法进行振动信号盲源分离的数值仿真,分离后的信号波形与源信号一致。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,不同故障的特征被分离开来。分析实例体现了所建立的方法在盲源信号分离中的有效性和对转子故障分离的实用性。

Blind Source Separation of Rotor Vibration Faults

The basic principle of blind source separation was introduced in the paper.Based upon the principle of minimum mutual information,a new method that estimates the separation matrix by independent component analysis was established.The method is named as speedup gradient method.Based on the method developed,the numerical simulation of vibration signal blind source separation was performed.The waveforms of separated signals are identical with source signals.This blind source separation method was applied to multi-sensor signals of a real rotor faults.The results show that the fault characteristics are separated satisfactorily.The real example shows the validity of this method and the practicability for rotors' faults separation.