基于时频滤波与EMD分解的单通道发动机振动信号盲源分离

摘要

由于传统盲源分离方法要求传感器观测信号数目不小于源信号数目,且在源信号平稳及相互独立的前提下,才能得到较为准确的分离信号,但在实际应用中,这些条件不易满足.针对发动机振动信号,提出一种基于时频滤波与EMD分解的单通道信号盲源分离方法,消除非平稳噪声的干扰,将通过EMD分解得到的本征模态分量作为虚拟的观测信号,并利用FastICA算法实现信号盲源分离.%As traditional blind source separation method needs more sensors to obtain measuring signals than the numbers of source signals and accurate separation signals can be achieved only if the source signals are stationary and independent,but such conditions can hardly be met in application,a blind source separation of engine's vibration,and single-channel source separation method based on time frequency filtering and empirical mode decomposition is proposed. The noise of non-stationary is filtered,and virtual measuring signals are derived from IMF. Separation signals are achieved using FastICA arithmetic.