基于局部特征尺度分解的旋转机械故障欠定盲源分离方法研究

针对传统盲源分离方法局限于源信号平稳、非高斯且相互独立,要求观测信号数不少于源信号数的问题,提出了一种基于局部特征尺度分解的旋转机械故障欠定盲源分离方法。采用局部特征尺度分解方法将观测信号分解为若干个内禀尺度分量,相当于对观测信号数进行升维,重组所有内禀尺度分量作为新的观测信号进行盲源分离,该方法不仅能分离非线性、非平稳的旋转机械故障信号,而且也可以解决观测信号数少于源信号数的欠定问题。通过模拟仿真和建立不平衡-碰摩-松动耦合故障的转子试验台进行试验分析,且与传统盲源分离方法进行对比,结果表明了该方法的有效性。

Underdetermined blind source separation method of rotating machinery faults based on local characteristic-scale decomposition

The traditional blind source separation is restricted that source signals should be non-gaussian, stationary and mutually independent and that the number of observations is assumed to be not less than the number of sources. Aiming at this problem, underdetermined blind source separation method of rotating machinery faults based on local characteristic-scale decomposition （LCD） is put forward. In the method, all the mechanical fault signals are decomposed into several intrinsic scale components （ISC） by LCD, so the number of observations become more. Then all the ISCs are composed into new observations, which were blindly separated. This method not only decompose non-linear and non-stationary signals of mechanical fault but also can solve the underdetermined problem that the number of observations is less than the number of sources. While combining that method with the traditional blind source separation, the analysis of simulation and experimental results of the rotor of unbalance-rubbing-loose- ness coupled faults show that this method is effective.