【作者】 张剑；

【导师】 刘昌文； 毕凤荣；

【作者基本信息】 天津大学 ， 动力机械及工程， 2016， 博士

【摘要】 爆震检测是汽油发动机点火闭环控制的前提和基础,爆震特征提取和强度评价是爆震检测与控制的前提。强烈爆震导致发动机功率下降,油耗上升,气缸过热甚至损坏发动机。但是,轻微爆震反而能改善发动机的动力性和效率。机体振动法是目前最为常用的爆震检测手段,但现有的爆震检测方法只能粗略的检测爆震,不能精细划分爆震强度,难于实现爆震的精确控制。因此,及时有效的检测爆震及判定爆震程度具有重要的理论意义和工程实用价值。论文在研究小波变换、经验模态分解、高阶统计量理论、分形理论和数学形态学等非平稳、非线性振动信号特征提取方法的基础上,系统深入地开展了信号分析方法在发动机振动信号爆震特征提取和强度判定领域中的应用研究,通过对测取信号的分析再处理,提取了表征发动机爆震状态的关键特征信息,实现了爆震检测和强度判定。主要研究成果如下:针对缸内压力高频振荡传播方向的随机性以及发动机机体结构振动的非线性,提出了融合多个方向振动特性分析爆震特征的方法,深入分析了缸内压力振荡对振动信号的影响规律,确定了对爆震最为敏感的振动方向。在深入研究EEMD和WPT基本原理和算法的基础上,针对爆震振动信号时频特性,结合EEMD滤波、WPT去噪的技术特点,提出了基于EEMD和WPT协同的爆震振动信号特征提取方法,克服了小波包分解频率成分无物理意义和IMF分量频带成分过宽的问题。提取了爆震特征子频带,通过局部能量分析提出爆震强度计算方法,能够有效地检测爆震并判定其强度。发动机处在不同强度爆震状态时,振动信号的非高斯成分在双频域内的分布强度与形态均不相同。通过对不同爆震强度振动信号的双谱特性分析,并利用双谱主对角切片和奇异值分解对双谱进行降维处理,有效实现了爆震特征提取和强度评价。在研究分形理论的基础上,探讨了盒维数和广义维数的计算方法。研究了单重分形和多重分形在爆震检测中的应用,对比了单重分形和多重分形检测爆震的优势与不足。为全面准确地识别爆震及其强度,提出了基于多域联合技术的特征提取方法:包括具有双重消噪能力的小波域双谱分析方法,突出爆震冲击特性的基于峭度准则和EEMD双谱的特征提取方法,准确描述爆震非高斯成分强度分布的双谱分形维数分析方法,量化爆震冲击的基于峭度准则和EEMD数学形态学分形的振动特性分析方法。这些方法从时频域、瞬时频率域、双频域以及数学形态几何等角度去认识爆震,揭示了新方法在发动机爆震特征提取及强度评价中的可行性和推广价值。更多还原

【Abstract】 Knock detection is the premise and foundation of the ignition closed-loop control in gasoline engine.Knock feature extraction and intensity evaluation is the premise of knock detection and control.Severe knock could make the power coastdown and fuel consumption increase,even damage mechanical parts while light knock can improve the power performance of an engine.Vibration detection technology is the most commonly used method of knock detection,but the existing knock detection method can not evaluate knock intensity accurately,so it is difficult to achieve precise control of the knock.Detecting knock timely and effectively has important theoretical significance and engineering practical value.This dissertation researched the non-stationary and non-linear vibration signal analysis and feature extraction method,including: Wavelet transform,empirical mode decomposition,higher order statistics theory,fractal theory and mathematical morphology,etc.Thesis researched signal analysis methods in the field of engine vibration signal analysis and detonation feature extraction and intensity evaluation.The main research results of the full text are as follows:In view of the randomness in the direction of propagation of cylinder pressure high-frequency oscillation and the non-linear vibration of engine structure block,this paper proposes a detonation characteristic analysis method based on the fusion of multiple directions vibration.The highly sensitive vibration direction of engine knock is proved by analyzing the influence rule of the in-cylinder pressure high-frequency oscillation on the vibration.On the basis of ensemble empirical mode decomposition & wavelet packet transformation principles and their calculation research,this paper works out signal abstraction method and further extracts the time-frequency characters and space distribution principles of vibration signals.The method,which overcomes the problem of the decomposition of wavelet packet frequency components without physical meaning and the too wide frequency band of the IMF component,can detect knock and evaluate the intensity effectively.When the engine knock is at different intensity,the distribution intensity and morphology of non-Gaussian components of vibration signals will change in bi-frequency domain along with the knock.Through the analysis of the characteristics of the vibration signal with different knock intensity,Knock feature extraction and intensity evaluation are effectively implemented by using bi-spectrum slice and singular value decomposition.The calculation method of box counting dimension and generalized dimension is discussed based on fractal geometry analysis.The single fractal and multi-fractal is studied in the application of knock detection.The fractal box dimension and generalized dimension of different knock conditions at different time are successfully extracted.This paper proposes a signal feature extraction method based on multi-domain united technologies,including dual de-noising ability of wavelet domain bi-spectrum analysis method,highlighted characteristic of knock impact based on kurtosis criterion and EEMD-Bispectrum feature extraction method,accurate description of knock non-Gaussian ingredients intensity distribution in bispectrum image based on fractal dimension analysis method,and quantization of based on kurtosis criterion& EEMD & mathematical morphology.These methods analyze knock vibration characteristic from different domain,such as time-frequency domain,instantaneous frequency domain,dual frequency domain and mathematical morphology geometrical.Derived from the above method,knock intensity factor has been developed and the relevance of the proposed criterion for characterizing different levels of knock has been investigated.The results reveal that the new method proposed is feasible in knock feature extraction and knock detection of gasoline engine.更多还原