| As one of the most widely used power plants, diesel engine is often applied to large and medium-sized cars and gradually applied to small cars and sedans. However, the severe vibration and noise problems restrict the use of diesel engines, especially when noise and vibration performance has received more attention. The identification of noise and vibration sources is of great significance to improve NVH performance of diesel engine. While diesel engine has a complex structure and works reciprocating to produce power, traditional noise source identification methods can only identify surface radiated noise sources and it’s hard to recognize inner physical noise sources. Recent years with the development of modern digital signal process technologies, the identification of inner physical noise sources has become a research hot topic in diesel engine industry.On the basis of previous studies, research on Time-Frequency Analysis and Blind Source Separation of diesel engine is carried out based on Empirical Mode Decomposition and Independent Component Analysis. EMD-RobustICA fusion method is proposed to identify the inner physical noise sources from single-channel complex noise signal.The novel method can conquer the problem that accurate separation can be achieved only if the numbers of sensors are more than the numbers of source signals. First the empirical mode decomposition of observed signal is carried out. Correlation analysis is used to eliminate mode mixing problems of EMD caused by second order filtering networking features. The rest of IMFS(Intrinsic Mode Function, IMF) after correlation analysis integrated with original observed signal are used as the inputs of RobustICA. Single channel observed noise signal independent component analysis could be achieved through RobustICA. No filtering operation is needed by means of anti-noise and robust performances of RobustICA. The separation result of simulated signal indicates the effectiveness of EMD-RobustICA fusion method proposed above.Applied to a four-cylinder-diesel-engine noise signal analysis, independent multiple sources hidden in the observed signal can be separated. Inner physical noise sources of diesel engine can be identified combined with FFT and CWT. Combustion noise, piston knock noise and fuel injection pump noise can be separated from the observed single channel noise signal successfully. Near field acoustic holography is used to identify the surface radiated noise sources of the diesel engine. Test result indicates that the main surface radiated noise sources are exhaust pipe, belt pulley and oil pan. Experiment results can provide theoretical support for the reduction of noise and vibration of diesel engine. |
