【作者】 徐红梅；

【导师】 郝志勇；

【作者基本信息】 浙江大学 ， 车辆工程， 2008， 博士

【摘要】 内燃机振声信号为典型的非平稳时变信号,频率成分十分复杂,采用传统的谱分析方法难以对其进行时域局部化分析,无法全面地揭示信号频率成分的时变特性。时频分析方法是一种时频局域化分析方法,能在时间和频率上同时描述信号的能量密度或强度,揭示信号中所包含的频率分量及其随时间变化的特性。近年来,随着数字信号处理技术的不断发展,基于非平稳信号处理的时频分析方法已逐渐应用于内燃机振声信号分析领域,并取得了一定的研究成果。本文主要围绕国家自然科学基金项目“内燃机声信号的独立成分小波分析及燃烧噪声源的诊断”和教育部高等学校博士学科点专项基金“通过发动机辐射声的小波分析诊断燃烧性能和噪声源”展开,在深入研究小波变换、S变换、Hilbert-Huang变换以及独立成分分析的基本原理与算法的基础上,以多台发动机为研究对象,系统深入地开展了时频分析方法和盲分离技术在内燃机振声信号时频特性分析和振动噪声源识别中的应用研究,研究结果对内燃机振动噪声控制具有重要意义。全文的主要研究内容如下:1、从噪声污染的严重性和危害性出发,阐述了开展内燃机振动噪声控制研究的实际意义;通过对小波变换、S变换、Hilbert-Huang变换以及独立成分分析的发展概况及其在工程领域的应用进行调查研究,分析了各种时频分析方法和盲分离技术应用于内燃机振声信号时频特性研究和振动噪声源识别的可行性。2、对时频分析的基本概念进行了定义与区分,归纳总结了时频分析的基本原理与基本性质。3、在对小波变换的基本原理进行深入研究的基础上,以多台发动机为研究对象,采用连续小波变换方法对其稳态和瞬态工况(加速过程)的振声信号进行时频分析,研究不同工况下振声信号的能量分布规律,以及其主要频率成分随时间(转速)变化的特性,结合内燃机的结构特点和工作机理,分析振动噪声产生的原因。4、采用小波包分解的方法对内燃机噪声信号进行小波包逐层分解和系数重构,将与内燃机燃烧过程相关的频带信号进行合成重构,得到燃烧激励产生的噪声信号,并对其进行连续复小波变换和三维小波能量谱分析,分析气缸燃烧状态,提取内燃机燃烧过程相关的特征信息。5、在对S变换基本原理和相应算法进行深入研究的基础上,开展S变换技术在内燃机振声信号分析中的应用研究。以多台发动机为研究对象,采用S变换对其稳态工况和瞬态工况(加速过程)的振声信号进行时频分析处理,分析振声信号的能量分布规律,以及其主要频率成分和能量分布随时间(转速)变化的情况,结合内燃机的结构特点,分析振动噪声产生的原因。6、在对Hilbert-Huang变换技术的基本原理和相关算法进行深入研究的基础上,开展Hilbert-Huang变换技术在内燃机振声信号分析中的应用研究。以某六缸发动机为研究对象,采用EMD方法对其振声信号进行分解,得到多个具有不同频率的IMF分量,分别对各分量进行Hilbert变换,分析其幅值和频率随时间变化的特性,并结合内燃机的结构和振动噪声产生机理,分析各分量产生的原因,识别振动噪声源。7、在对独立成分分析的基本理论和相关算法进行深入研究的基础上,开展独立成分分析在内燃机振声信号分析中的应用研究。通过对内燃机振声信号的独立性和高斯性进行分析,揭示ICA方法在内燃机振声信号分析中的可行性,并以某六缸发动机为研究对象,采用ICA方法分别对其振声信号进行盲分离,结合小波变换技术识别内燃机主要噪声源和机体的振动激励源。更多还原

【Abstract】 The vibration and noise signal of internal combustion (I.C.) engine is typically non-stationary and time-varying with complicated frequency components. Traditional spectral analysis methods can neither analyze this signal in local time domain nor reflect the time-varying characteristics of the frequency components. The time-frequency analysis method is characterized by time-frequency localization. It can simultaneously describe the energy density or intensity of signal in time and frequency. Therefore, this method can reveal the frequency components included in signal and the related variation with time. In recent years, with the development of digital signal processing technologies, the time-frequency analysis method suitable for processing non-stationary signals, has gradually been introduced and brought on some achievements in processing the vibration and noise signal of I.C. engine.This research is mainly originated from the project "Diagnosis of combustion noise source based on independent component analysis and wavelet transform of engine sound signal"(National Natural Science Foundation Project, No:50575203) and "Diagnosis of combustion performance and noise source based on wavelet analysis of engine radiation noise"(ph.D. Programs Foundation Project of Ministry of Education, No: 20030335092). The basic principles and algorithms of wavelet transform, S transform, Hilbert-Huang transform together with independent component analysis were investigated deeply and thoroughly. Taking several engines for example, the time-frequency analysis method and blind source separation technique were employed to study on the time and frequency characteristics of engine vibration and noise signals as well as to identify the vibration and noise sources. The research results are of great significance to the vibration and noise control of engine. The research details were as follows:1. Starting with the severity and harmfulness of noise, the importance of controlling engine vibration and noise was then expatiated. After investigating the development and applications of wavelet transform, S transform, Hilbert-Huang transform, and independent component analysis, the feasibility of these techniques used for analyzing the time and frequency characteristics and identifying vibration and noise sources of engine was explored.2.The basic concepts of time-frequency analysis were firstly defined and differentiated. Then the basic principles and characteristics of time-frequency analysis were summarized.3. Based on an in-depth study of the basic principles of wavelet transform, continuous wavelet transform was adopted to analyze the vibration and noise signals in time-frequency domain when engine was under steady and transient workload. The energy distribution of signals recorded in different conditions and variance of their predominant frequency components with time or engine speed were subsequently investigated. Afterwards, the cause of engine vibration and noise was analyzed according to the previous results together with the engine structural characteristics and working mechanism.4. Based on the wavelet packet decomposition and coefficient reconstruction of engine noise signal, the subband signals relevant to engine combustion process were combined and reconstructed, and noise signals resulted from the combustion excitation were obtained. Moreover, continuous complex wavelet transform was adopted to analyze the reconstructed signal. The three-dimensional wavelet power spectrum was also calculated to investigate the combustion condition of each cylinder and extract the related information.5. Based on an in-depth study of the basic principles and relevant algorithms of S transform, the research on the application of S transform in analysis of engine vibration and noise signals was conducted. Taking several engines for example, S transform was employed to process vibration and noise signals in time-frequency domain when engine was under steady and transient workload. Then the energy distribution law of signals was studied. Furthermore, variation of prominent frequency components with time was analyzed. Finally, the cause of vibration and noise was deduced according to engine structural characteristics.6. Based on an in-depth study of the basic principles and relevant algorithms of Hilbert-Huang transform, the research on its application in analysis of engine vibration and noise signals was conducted. Taking a six cylinder engine for example, the vibration and noise signals were decomposed into a collection of intrinsic mode functions (IMFs) with different frequencies by the method of EMD. Then Hilbert transform was adopted to analyze the amplitude and frequency of each IMF varying with time. In terms of the mechanism of the engine structural vibration and noise, the origination of each IMF was deduced. The vibration and noise sources were identified as well.7. Based on an in-depth study of the basic principles and relevant algorithms of independent component analysis, the research on its application in analysis of engine vibration and noise signals was conducted. The independence and gaussianity of vibration and noise signals was firstly investigated to demonstrate the feasibility of this method used for analyzing the vibration and noise signals of engine. Taking a six cylinder engine for example, the technology of independent component analysis and wavelet transform was adopted to separate the vibration and noise signals of engine and identify its vibration and noise sources.更多还原