基于变分模态分解和奇异值分解的频率相近信号分离方法

频率相近信号分离是故障诊断的难点问题。变分模态分解(VMD)作为一种新的信号时频分析方法,对频率相近信号具有较高的分辨率,能够实现频率相近信号的分离。VMD分解时先指定分解层数,层数选取的优劣将直接影响分解效果,过分解时容易产生虚假频率成分,反之,欠分解则容易丢失有用频率成分。基于此,提出一种基于变分模态分解和奇异值分解相结合的相近频率信号分离方法。首先选择适当的分解层数对信号进行VMD过分解,然后对分解得到的分量进行奇异值分解,通过奇异值分解检测并剔除虚假信号成分,从而实现频率相近信号的有效分离。利用仿真信号和滚动轴承故障信号证明了该方法的有效性和可行性。

Similar Frequency Signal Separation Based on VMD and Singular Value Decomposition

Similar frequency signal separation is a difficult problem in fault diagnosis. As a new method of signal time frequency analysis, variational mode decomposition (VMD) has a higher resolution for signals with similar frequency. The number of decomposition levels, which can directly affect the decomposability, is first specified in VMD. Once over-decompose is likely to produce false frequency components, while under-decompose is easy to lose useful frequency components. Thus, a new method of similar frequency signal separation based on VMD and signal singular value decomposition is proposed. Firstly, appropriate decomposition levels is selected to over-decompose the signal, and then singular value decomposition is carried out on the components obtained by VMD, which can detect and eliminate false signal components, so as to separate similar frequency signal well. The effectiveness and feasibility of the proposed method are demonstrated by simulation signal and rolling bearing fault signal.