Research On Technology Of Blind Source Separation Based On Surface Radiation Noise Of I. C. Engine

The study on noise and vibration of internal combustion engine were paid more attention rencently by internal researchers for the improving requirement of vehicles′NVH (noise vibration and harshness) performance. The identifications of the main noise source and the research on mechanism of engine noise emission are also critical for both noise control engineer and fault diagnose. The vibration and noise emission from engine is complicated for there are too many internal excitations of engine, and the excitations are also time variable signals from I.C. engine show their limitation in describing the complexity. The digital signal processing (DSP) technique had got great achievement in recent twenty years.Base on the characteristics of internal-combustion engine noise signal, the noise test measurement, wavelet transform, independent component analysis and eliminating blind uncertainty of ICA estimation to the research of vibration and noise signals from I.C. engine was deeply investigated. The main noise source of the engine and the contribution are separated and identified from the complex noise. The main research contents are as follows:1. The mechanism of internal-combustion engine noise was expounded. The main excitations and the path of noise transmission in the internal parts of the engine were revealed. The noise and vibration signal was no-stationary induced by both the time-variable excitations and time-dependent transmission path. The methods of identification of internal combustion engine were reviewed.2. Basic on the principle of independent component analysis. The statistical independence and guassianity of noise signals at different operatation conditions of a four cylinder diesel engine was analysed. The result has shown that noise signals meet the preconditions of ICA. The negative entropy enormously fast independent component analysis (FastICA) algorithm was adopted to separate noise signals, and a series of independent components were obtained.3. In order to identify eath independent component further, combining with the feature of time-frequency analysis results and the mechanism of the engine radiation noise, the FFT and the wavelet transform are used to determine the relationship between different independent components and different noise source of the engine.4. In order to eliminate blind uncertainty of ICA estimation, a adaptive method for revision of ICA estimate was adopted bsed on combination of FFT-MCC. Thus, blind uncertainty of ICA was eliminated and the waveforms of sources were restored correctly.5. Through the calculation of the sound power radiated of the separated singnal from noise signal of independent component analysis and the whole engine, the contribution of noise source and characteristics were obtained. The main noise source control measures were put forward as a foundation to control the whole engine noise. Experimental results verified the correction of the research about the internal combustion engine noise blind source separation technology.