摘要
对独立成分分析的基本原理和数学模型进行了简要叙述,以某六缸柴油机为研究对象,对其不同工况下的噪声信号进行了统计独立性和高斯性分析,噪声信号基本满足独立成分分析的前提条件。采用基于峭度的梯度算法对噪声信号进行了盲分离,得到一序列独立分量。为进一步识别各独立分量,采用傅立叶和连续复小波变换对其进行时频分析,并结合一些内燃机先验知识分析发现,这些独立分量基本上对应着内燃机的燃烧噪声、正时齿轮噪声、活塞敲击噪声等噪声源,因此,采用独立成分小波分析技术对内燃机噪声信号进行盲分离以识别其主要噪声源是可行的。
The basic principles and mathematical model of independent component analysis （ICA） were briefly introduced. The statistical independence and gaussianity of noise signals at different operation conditions of a six cylinder diesel engine was analyzed. The result has shown that noise signals meet the preconditions of ICA. The gradient algorithm based on the kurtosis was adopted to separate noise signals, and a series of independent components were obtained. In order to identify each independent component further, Fourier and continuous complex wavelet transform was adopted to analyze time-frequency domain of these components. In combination with some prior information of internal combustion engine, the analysis found that these independent components basically correspond to the major noise sources of engine, such as combustion noise, timing gear noise, piston slapping noise, and so on. It＇s feasible to adopt the technology of independent component and wavelet analysis to separate the engine noise signals and to identify major noise sources.
出处
《内燃机工程》
EI
CAS
CSCD
北大核心
2007年第6期61-65,共5页
Chinese Internal Combustion Engine Engineering
基金
国家自然科学基金资助项目(50575203)
关键词
内燃机
噪声源识别
独立成分分析
小波分析
盲分离
IC engine
noise source identification
independent component analysis
wavelet transform
blind separation
