Study On Blind Source Separation Of Radiation Noise Of Internal Combustion Engine

The performance of noise,vibration and harshness(NVH)of avehicle is required to improve and is attracting more attention by customers.As one of the main noise sources of traditional vehicles,the noise from the internal combustion engine is considered to be the most importantin NVH performance improvement for a vehicle.Separation and identification of contribution sources from the noise of the internal combustion engine are keys to control the total level of engine radiation noise.The excitation diversity of the engine noise sourcesdepends on the enginestructure complexity.Traditional engine noise identification methods have certain limitations in practical applications.However,the development of modern signal processing technology provides new ways for engine noise signal analysis and processing.This thesis focuses on how to effectively separate and identify the excitation sourcesfor an internal combustion engine.The research work for the noise of a single cylinder gasoline enginein this thesis includes noise generation mechanism study,algorithm development of blind source signal processing and performing noise separation and identification for the engine based on the signal processing.Using the modern signal processing technology to process the engine radiation noise,the engine main noise sourceshave been separated and identified and they are useful to help control the total level of engine radiation noise.The content of this thesis includes:(1)Research on mechanism of surface radiation noise of internal combustion engine.The main noise source of internal combustion engineand its mechanism are analyzed in details,and the source of noise and its main transmission path are described.The radiation noise signal of internal combustion engine has characters of non-stability and periodicity.Separation and identification methods for the engine noise are summarized from two aspects,the traditional noise identification method and the modern signal processing technology.(2)Research on method of independent component analysis in signal processing algorithm.Established the mathematical model for independent component analysis and pointed out constraint conditions for algorithm applications.Based on the independent component analysis criteria,mostly applied methods of independent component analysis are summarized in the thesis.(3)Research on applications of FastICA algorithm to general analog signals and noise signals of a single cylinder engine.Based on the calculation procedure of FastICA algorithm,the method is applied to the separation of analog signals to verify the feasibility of the method.Through the analysis of the collected noise signals to meet application conditions,FastICA algorithm is then applied to separate noise signals and five independent noise components are obtained.(4)Research on noise identifications of a single cylinder engine.The common methods of Fourier transform and wavelet transform in time-frequency domain analysis are applied to process analog signals and to verify the feasibility of wavelet transform.The wavelet transform is applied to analyze independent source signals of the single cylinder engine.Through the analysis of specific data of each independent component and combination with the theory of frequency characters of the main noise source of engine,the independent components respectively correspond to combustion noise,piston knocking noise,injection pump noise and chain gear noise.Contributionsfrom the above components to the total noise are 43.64%,24.65%,10.41% and 7.88%,respectively,by acoustic summaries.