Study On The Separation Of Diesel Engine Noise Based On Independent Component Analysis

As a typical form of vehicle dynamic noise, diesel engine pollution has brought a series of mental and physical harm to urban people. The restriction regulation against diesel engine noises pollution become more stringent all over the world as long as people's environmental consciousnesses are rising. Because noise is such an important index of vehicle's overall performance which attracts lots of attention by automobile manufacturers and consumers, many automobile manufacturers and their R&D team treat engine noise, vibration, and smooth operation as the focus. The fully illustrated examples mentioned above have proved the importance of Noise Control to diesel engines.The principle of diesel engine noise generation is so complicated that it is hard to use traditional methods to separate noise signals completely. A period of time, noise separation problem is like a bottleneck which has been restricted engine noise research, analysis, management. Noises produced from a working diesel engine are typical blind source signals which could be separated by independent component analysis.Independent component analysis is an effective method which has been widely used to deal with blind source separation cases. It can take full advantage of statistical independence assumption and non-Gaussian assumption, and by constitute an objective function then optimize it so the independent component can be taken from the observation data. This article describes the FastICA algorithm, which has the advantage of fast convergence and calculation, is commonly used in the treatment algorithm for engineering cases. Results of MATLAB simulation test show that the waveforms and temporal information belong to various components perfectly match. The simulation result is satisfactory, but also proved that independent component analysis is a high reliability method.According to national standards, the testing diesel engine noise signals under 4 conditions are collected precisely. Independent components are separated successfully after the algorithm application necessary conditions. It is shown that components independent with each other are originated from piston knocking, pilot-injection and combustion respectively. Test results provided a valuable application basis for the task of diesel engine noise control.In this article, independent component analysis is used to research on identification and separation work to a diesel engine noise. Beyond one-dimensional analysis, Morlet wavelet can be used as a wavelet basis function to do research on time-frequency analysis under various working conditions. Based on wavelet transform technique, the two-dimensional analysis which is focus on time and frequency can offer a more comprehensive description. Besides, it can further distinguish various noise signals according to two-dimension information.