基于VMD和ICA的发动机故障特征增强研究

Research on Enhancement Method Based on VMD and ICA for Engine Fault Feature

针对发动机噪声信号信噪比低、故障特征提取困难等问题,提出一种基于变分模态分解(VMD)结合独立分量分析(ICA)的降噪方法。首先通过VMD分解得到分量中心频率确定合适的分解层数,以峭度准则重构噪声信号和故障信号,然后通过FastICA将重构信号再次分解,得到降噪后的故障信号,最后与经验模态分解(EMD)降噪对比,结合实例和仿真表明,该方法能够有效抑制模态混叠,增强发动机噪声信号故障特征。

To solve the engine noise signal problems such as low signal-to-noise ratio and difficult extraction of fault features,the paper puts forward a new noise reduction method on the basis of variational mode decomposition(VMD)and independent component analysis(ICA).This method determines the proper decomposition layers by component center frequency decomposed with VMD,uses kurtosis criterion to reconstruct the noise and fault signals,which are decomposed again with FastICA to obtain the denoised fault signal.Both example and simulation results show that when compared to empirical mode decomposition(EMD)noise reduction,this method can effectively suppress modal overlap and enhance fault features.