NOISE SOURCE SEPARATION OF DIESEL ENGINE BASED ON EMPIRICAL WAVELET TRANSFORM AND ROBUSTICA METHOD

摘要

Through separating the noise sources of diesel engine, the obtained independent noise source can provide basis for the vibration noise reduction and monitoring diagnosis. In the separation of noise source, the combustion noise and piston slap noise are seriously aliased in time domain and frequency domain. In order to separate them, the EWT-RobustICA method is proposed. The vibration and noise test of diesel engine is designed. In order to shield the other cylinder interference noise, the outer surfaces of cylinder body are carried out the silencer cotton and lead cover treatment, with only No. 6 cylinder parts exposed. Firstly, the EWT method is used to decompose the measured single-channel noise signal. Then the obtained empirical mode components which have a high correlation coefficient with the measured noise signal are selected. The selected empirical mode components and the measured noise signal are put together to form a new signal group. Finally, the RobustICA algorithm is utilized to extract independent components. The separated results are further identified by FFT, continuous wavelet transform, coherent analysis and back-drag test. The experimental results show that the proposed EWT-RobustICA method can obtain more accurate and pure combustion noise and piston slap noise than EMD-RobustICA method. The frequency of combustion noise and piston slap noise are concentrated at 3813 Hz and 1675 Hz, respectively.