EMD-Robust ICA在柴油机噪声源识别中的应用

Application of EMD-Robust ICA to Noise Sources Identification of Diesel Engines

为了有效地从复杂的单一通道噪声信号中分离和识别柴油机的噪声源,采用经验模态分解(EMD)和基于峭度的鲁棒性独立分量分析(RobustICA)相结合的方法,将EMD分解后的本征模态函数与原噪声信号作为RobustICA的输入,借助RobustICA良好的抗噪性,不需要对观测信号进行滤波处理就可以实现单一通道观测信号的源分量分离。模拟仿真的结果充分说明了该方法的可行性。应用于某四缸柴油机噪声信号分析,对分离出的独立分量进行小波(CWT)时频分析,结合内燃机的特性,从单一通道噪声信号中准确识别出柴油机的燃烧噪声和活塞敲击噪声。

To effectively identify the noise sources of diesel engines from single- channel noise signal, the methodcombining the empirical mode decomposition （EMD） with the robust independent component analysis （Robust ICA） isproposed. In this method, the instinct mode functions （IMF） after correlation analysis integrated with original noise signalare used as virtual channels of Robust ICA. In virtue of the anti-noise advantage of Robust ICA, the single-channel noisesignal can be separated into several originally independent components without filtering processing. Result of simulationshows the effectiveness and feasibility of this method. Then, this method is applied to the noise signal analysis of a fourcylinderdiesel engine. The continuous wavelet transform （CWT） is used for analysis of the separated independentcomponents in time and frequency domains. Combined with fast Fourier transform （FFT）, the combustion noise and pistonslap noise of the diesel engine can be identified precisely from the single-channel noise signal.