Application Of EMD And ICA To Vibration And Acoustic Signal Processing In Internal Combustion Engine

Internal combustion engine has a complex structure, and works randomly andchangeably. The vibration and noise it produced has typical impact properties, thusthey are unsteady time-varying signal. Traditional signal analysis methods can onlyanalysis signal’s statistical average characteristic, while the pattern that signalfrequency changes with time cannot be fully revealed. Time-frequency analysis canprovide panorama or localization attribute and intrinsic properties of signal in bothtime domain and frequency domain, and then frequency component of the signalwhich varies with time can be found. In recent years, with the rapid development ofsignal processing technology, time-frequency analysis is commonly used to deal withunsteady signal, especially vibration and noise signal in power machinery.This research is mainly originated from the project Research on Vibration andNoise Control of Engine supported by National High-tech Research and DevelopmentProgram (863Torch Program). Research on application of time-frequency analysismethod and blind source separation technique on internal combustion engine iscarried out based on Hilbert-Huang transform and independent component analysis.The analysis results are to be used as foundation of vibration and noise control ofinternal combustion engine.The main research work is as follows:(1)Experimental data acquisition: Capturing vibration and noise signals of adiesel engine under different working conditions by using LMS data acquisitionsystem.(2)Introduce the theory of Hilbert-Huang transform, and its application in theunsteady vibration and noise signal. Hilbert-Huang transform is used to process thesignals collected from cylinder head of six cylinder four stroke water-cooled dieselengine. Firstly, the vibration signal is decompose by empirical mode decomposition(EMD) decomposition). Thus intrinsic mode functions (IMF components) withdifferent frequencies are obtained. Each IMF components is transformed by usingHilbert transform, then the intrinsic properties of signal in both time domain andfrequency domain is analyzed. Finally, each IMF component is researched withvibration mechanism of internal combustion engine and priori knowledge. The results are likely to be of significance in controlling the vibration.(3)Application of EMD-ICA in noise source identification in internal combustionengine is studied. Intrinsic mode functions got from the empirical modedecomposition are used as multiple virtual channels of independent componentanalysis. Therefore, the underdetermined problem of source identification from singlesampling signal is solved. This approach is used for noise source identification ofmain radiated noise in internal combustion engine, combined with correlation analysisand wavelet time-frequency analysis. The results show that EMD-ICA is effective innoise source identification, and overcomes the modal aliasing problems of the EMDtechnique in noise source identification. And difficulty is reduced when ICA isadopted to identify noise sources with single sampling signal.(4)Signal processing software is programmed by using MATLAB andLABVIEW.