Identification of Diesel Sound Source Based on The Independent Component Analysis

As a new approach of blind source separation (BSS), independent component analysis (ICA) has attracted extensive attention of researchers in the field of information processing. In this paper, the basic theory and algorithm of ICA are briefly introduced, and then ICA is used for the preprocessing of engine acoustic signals to identify the engine noise sources. The ICA decomposes the signals into a number of independent components (ICs) so the individual engine acoustic sources can be studied separately. The theoretical description of the characteristics of the diesel engine noise sources is introduced. A numerical example is presented to verify the separation efficiency of the ICA. The numerical example shows that the ICA can effectively separate the embedded low-level sources. Although acoustic signals from diesel engines contain useful information, it is still difficult to resolve them accurately based on noise measurement. This is because the occurrences of each noise source are too close together. The low-level noise sources cannot be indentified successfully. The ICA model is used to extract the above engine noise sources of two types of engines. The continuous wavelet transform (CWT) is applied to represent the ICs in the time-frequency domain. The source separation results from the recorded acoustic signals are accepted.